-*- change-log -*-
Started on CPAN with B-C-1.04_12
The Perl compiler was in CORE from alpha4 until Perl 5.9.4
and worked quite fine with Perl 5.6 and 5.8.
1.55 2016-05-?? rurban
Support 5.24 and 5.24c
* C: handle cperl OP_SIGNATURE
Fix refcount of cop hints hashes with 5.22-nt (#220)
Add missing B::FAKEOP::rettype for cperl-5.24, causing SEGV on B::CC
Fixed shared xpviv with 5.24 on 32bit.
Fix various compiler warnings in ByteLoader
Fix bytecode.pl call in cperl core.
1.54_03 2016-05-09 rurban
Released with cperl-5.22.2
* C: Remove hash argument from share_hek, always rehash.
Work on PERL_DESTRUCT_LEVEL=2 Unbalanced string table refcount issues.
Fix for CVf_CONST CVs with cperl (constant::import)
Replace %Lu with %lu/%u
Fix for shared xpviv with 5.24. This asserts now if unshared.
* perlcc/cc_harness/TestBC: many more PERL_CORE fixes, to workaround
broken ExtUtils::Embed::ldopts
* tests: skip perlcc with -m32 cross and PERL_CORE. Can only work reliable
on individual CPAN setups, but not with CORE smoking. Blame Configure
Replace test.pl with TestBC.pm. Blame core harness
1.54 2016-02-26 rurban
cperl fixes and 5.22 memory improvements. cperl now also supported,
even included CORE, but ByteLoader still broken.
* C: Fix defined for RV since 5.20 and ithreads (#354)
Fix anon functions in INIT block (#352) (atoomic)
and more const CV fixes.
Fix C.xs for old non-c99 compilers. i.e. msvc6
Fix segfault from PVMG (#348) (atoomic), regression added
for match once
Fix stack corruption in HV::ENAMES (#351)
Fix SEGV with empty get_cv(), esp. from non-XS functions (#358)
Fixed 5.22 padname length limitation of max 60. Also using less memory. (#361)
New -O2 option -fcow since 5.20 to cow most static strings with ~6%
memory savings (#361)
Support cperl5.22.2 with AvSTATIC and AvIsCOW. ~6% less memory. (#361)
Support cperl5.22.2 with HEK_STATIC. ~6% less memory. (#361)
cperl fixes for DynaLoader, XSLoader improvements (#363), -O3 destruction.
* B::C::Flags: renamed to B::C::Config
* bytecode.pl: support cperl bootstrapping with miniperl in CORE
* Asmdata (1.04): with cperl in CORE @optype @specialsv_name are defined here
also. Cannot load B with miniperl.
1.53 2015-12-06 rurban
added 5.22 support, some cperl optimizations and fixed most remaining 5.14 issues.
Bytecode broken upstream with 5.22.0
* Bytecode (1.17): removed lastsib bit for 5.21.2-5.21.10, support only moresib op
Add ops for padl, padn, padnl, unop_aux and methop needed for 5.22
ByteLoader was broken upstream with 5.22. You need to build a perl with
`perlall build --patches-Compiler 5.22.0` or use cperl5.22.2.
Added a probe, set $B::C::Flags::have_byteloader
* Disassembler (1.13), Assembler (1.13): support PADNAME and PADNAMELIST for 5.22
* C: fix B::PMOP::precomp for UTF8 (#333, #338, GH#50)
Add 5.22 support with a new unop_aux (multideref) section, methop support,
new support for PADNAME and PADNAMELIST objects.
Fix upstream B::UNOP_AUX::aux_list bug with threads. Use our own method which
returns the PADOFFSET, not the SV. (#341)
Removed B::Section which used base. (toddr)
Re-add PL_sv_objcount on cperl
Fixed XSUB CONSTSUB alias (44,45,... PR #228) (atoomic)
and anonymous CVs for 5.22 (#246, #305)
Proper utf8 support for gv_fetchpv, gv_stashpv and cv_get,
cache gv_stashpv calls.
Fix 5.22 for empty hash keys, share_hek("") (272, GH#249)
Fixed refcounts for shared hek's, fixing the notorious unshare_hek assertions
since 5.10 (GH #251, #255). See 1.06: "Had to disable -O1 -fcog (pv_copy_on_grow)
on 5.10 and higher until I find out how to fool S_unshare_hek_or_pvn. This fixes
all C -O1 and -O2 tests. Warn about this."
Fix Inf/NaN support for C (GH#287)
Fix handling of READONLY hashes in 5.22, i.e. use feature (GH#250)
Support custom op Devel_Peek_Dump, added with 5.20 (GH#274)
Defer regex compilation to new init1 for SWASHNEW (GH#273)
Support unicode labels (upstream B bug) (GH#318)
Skip unneeded init-time lexwarn checks,
support optimized cperl PERL_SUPPORT_STATIC_COP (cperl #70)
Fixed overload stringify broken since 5.18 (GH #219)
by adding a mro_isa_changed_in() to the overloaded stash.
Support multiple stash ENAMES, needed since 5.14 (GH #331)
This fixed many old 5.14 limitations. (Added to our version of B)
Support %^H, storing the hints_hash (features, pragmas) (GH#220)
This fixed many old 5.14 limitations.
Improve MATCH once (m??) handling with reset, find PMOP from
the Regexp dynamically (GH#252)
Bump refcounts of dynamic get_cv refs (#293). Fixes Coro in global destruction.
Skip more XS defined POSIX constants: all that start with 'M' (GH#335, #345)
One more "Eval-group not allowed at runtime" fix (GH#137, GH#346)
* CC (1.16_01): Encode unicode labels (GH#318)
* Stackobj (1.12_01): fix Inf/NaN support for CC (GH#287)
* t/CORE: versioned the core tests and moved to a external submodule
perl11/p5-coretests on github. (GH#332)
1.52 2014-09-09 rurban
* C: Protect against empty SV ptr in SV magic (\0) with $` with 5.20 (#370)
Save new pmop (?{}) code_list with 5.18 (#372)
Unset PL_use_safe_putenv as in perlmain, which initializes PL_origalen properly,
which fixes setting long a $0 string (#194)
Force Moose when only Class::MOP is loaded to give the needed path hint to XSLoader,
when "Assuming xs loaded $stashname".
* t/test.pl: print @cmd with TEST_VERBOSE
1.51 2014-08-04 rurban
* C: boot_EV (analog to Coro) SvREADONLY_off its XS symbols (#368)
1.50 2014-07-23 rurban
* C: fix DBI, special case DBI_MAGIC, call DBI->_install_method to
re-initialize compile-time internal ima pointers (#359)
Skip saving %IO::Handle for *STDOUT, which bloated 5.18 (#361)
Fix XS boot of Class::MOP without Moose, and generalize it to other
xs-loaded modules, which just call bootstrap Module. (#350, #364)
Earlier xs_init for Encode similar to attributes (hack, fixes #32 regression)
Use the system malloc for data being system free'd: cop_warnings,
PMOP->op_pv, sv_debug_file, needed for certain DEBUGGING options.
Fixes Attribute::Handler free errors (#362)
Strip binary BM chunk from comppad names < 5.8.8 (#363)
Save some CvXSUBANY values, esp. relocate any_ptr if possible:
POSIX::is*, DBI, List::MoreUtils::*_iterator.
The any_i32 values are already set in the boot section of the XS.
Fixed initialization of foreign Encodings (#365),
fixed initialization of old Encode versions < 2.58 also (#305)
* CC (1.16): Strip binary BM chunk from comppad names (CC names) < 5.8.8 (#363)
* perlcc (2.20): Strip wrong version-less -I<site_perl> (#366)
* Makefile.PL: Install into archlib on 5.8 and 5.6 (#358)
1.49 2014-07-12 rurban
* C: Changed IO::Socket::SSL warning to recommend >= 1.995.
1.996 even has some perlcc specific usage documentation (#317)
Fix PerlIO::scalar XS bootstrap forced by __DATA__ (#360)
We sometimes missed it
1.48 2014-07-11 rurban
Fixed C walker and 5.18-5.20 Bytecode.
This is the major version which cPanel will use for 5.14.4. With the next
versions we'll start working on -m and buildcc, to split packages into shared libs.
And use B::CC and/or rperl compiled modules selectively.
* C: Make revised -fwalkall the default. Fixes missing dumped method-only packages (#348)
Always save all user-loaded packages. Skip only compiler-loaded packages if
not used by user code. The compiler loads and dumps now additionally only the
heavy parts for utf8 and bytes, and utf8 for m///i, and AnyDBM for dbmopen,
PerlIO::Scalar for __DATA__ handles, and Coro::State::_jit if required, but
nothing else.
Fix for missing RTLD_NOLOAD on BSD, needed for run-time remap of compile-time
XS symbols via dlsym (#351)
Special-case XS loading of Moose.xs from Class::MOP (#350)
Better fix for mult. match once by checking OP_MATCH with ONCE flags in C.xs also (#274)
but there are still some platforms without pmop->op_pmregexp in the C.xs detection runloop
Add op_lastsib bit for 5.21.2
Changed IO::Socket::SSL warning as our patch was rejected upstream.
Need to maintain now a fork cPanel::IO::Socket::SSL to be able to use compiled
IO::Socket::SSL servers. (#317)
* Stackobj (1.12): Harmonize RV checks (fixes cc 105 -O1 and -O2)
* Bytecode (1.16): Replace each %hash with foreach keys in walksymtable also (#307)
Do not store main_cv->START, empty optree since 5.18
Add op_lastsib for 5.21.2, but sort still broken with PERL_OP_PARENT
* Assembler (1.12): error if a PUT method misses an argument
* perlcc (2.19): -B produces again -H bytecode, runnable without -MByteLoader.
Fix default output name from a.outexe to a.exe on a windows-based OS.
* Makefile.PL: fix installation of cc_runtime.h for B::CC
1.47 2014-06-11 rurban
Fixed Bytecode for 5.18. Fully support 5.18 and 5.20 now.
Greatly reduce code size for C and esp. CC, by including less unneeded dependencies.
* C: detect new Encode >2.58 compile-time usage (#305, RT #94221)
Fix mro maybe::next:: detection (#326)
Fix %INC cleanup logic. Delete packages from %INC only if they were really not saved,
and add packages to %INC which were saved and not just marked for saving. (#340)
Fix lot of skip package logic and consistency parts. Much less packages are now saved
in general which leads to smaller and faster code.
Avoid compile-time compiler internal reloads and redefinitions.
Special-case %warning::Bits on -O3 to not SEGV when changing warnings on run-time loads
on compile-time saved -fconst-string warnings.
Add experimental -fwalkall to add all wanted packages recursively with maxdepth 3.
No tests yet.
Add time and version on top of C source code. Also for CC.
Fix lexical subs for threaded perls (#130, #341)
Add special fixes for Coro: reload Coro jit and SvREADONLY_off some of its XS symbols (#293)
Fix "Eval-group not allowed at runtime" errors since 5.18 (#137),
by settting HINT_RE_EVAL for re-eval groups when compiling the QR
Fix for readonly magic hashes, usually found since 5.20 (#273)
Fix Can't locate object method "bootstrap_inherit" via package "DynaLoader" (#125)
again, as the improved walker now strips DynaLoader from simple one-liners like
'use Clone' and the DynaLoader detection was too late. Do it now earlier.
* ByteLoader (0.11): fixed filter -H problem with 5.18 (#339)
Disabling the PerlIO_binmode(PL_RSFP, IoTYPE_RDONLY, O_BINARY, 0); hack for [perl #86186]
"Changing filters to be textmode, not binary" fixed the Bytecode problems since 5.18
* perlcc (2.18): Avoid -H for bytecode > 5.18
New --check option
* t/asmdata.t: signicant enhancements
* t/CORE: add mro and re core tests, only mro/isarev.t is unstable, and several re-eval
tests using variables are skipped and unsupported.
1.46 2014-05-14 rurban
Added global destruction, improved -O4 and CC, fixed mro, @-, $/ $\, Net::DNS, ...
Added mro and re core testsuite.
* C: reset PL_stack_sp on init (was one too far). no effects on code.
Implement global destruction via sv_clean_objs and my_curse on -O3, resp.
prepend static svs PL_sv_arenaroot for sv_clean_objs with a fake sv_list[0].
(#197, #208, #254, #280)
Call DESTROY methods for all our global static SVs (i.e. "global destruction"),
fixes #197, #280, #282
Increment GV->FORM refcounts as they cannot be destructed, analog to CV
Call PL_exitlist functions with -O3
Make the lexical cop_warnings pointer dynamic, because on cv_undef (scope exit,
assign, exit, die, ...) CvROOT and all its kids are freed, ignoring op_latefreed.
The content of the warning is still a static string - and leaks on perl, but
we don't care.
-fwarn-sv is now always set, just not with MSVC on Windows. The warnings are then
directly set, not in a tight loop afterwards, because the value is needed for
the dynamic init.
Fixed lexical warnings on 5.8 (pWARN_STD .. pWARN_NONE)
Improve -O4 by keeping all CvSTART cops
Keep internal packages if used in the source code, e.g. mro (#300)
Fix the order of PerlIO_cleanup, after global destruction (#302, #303)
Since 5.18 padlists are now all dynamic, due to undef issues in non-local exits (#304)
Add -fno-dyn-padlist to -O4 to keep static padlists on 5.18.
Dynamic padlists are needed to prevent from cv_undef crashes on static padlists
when cleaning up the stack on non-local exits, like die or exit or subs on the
compile stack as with Attribute::Handler (#169, #304). Previously only END block
function padlists were dynamic (#298).
Add run-time remap of compile-time XS symbols via dlsym (#305, RT #94069)
Net::DNS includes Encode::XS ascii_encoding, but also init the 3 other encoding ptrs.
Add a fast path for d_dlopen + i_dlfcn systems (i.e. Windows goes the slow init2 route)
Detect new Encode-2.58 which stores encodings better to handle proper
compile-time usage (#305, RT #94221)
Add special SV associations for $/, $@ and $\ (#306, #256)
Save @ISA for all included packages, esp. DynaLoader dependencies (#308)
Replace each %hash with foreach sort keys (#307)
Set ${^GLOBAL_PHASE} with -O3 for END and DESTRUCT (#197, #208)
Do not store invalid GvCV symbols, only cv or &sv_list entries (#313)
Support GvFORM pseudo CVs (B::FM objects) on 5.10 (#149)
Support local $/ = "somestring" (only used empty values before) (#314),
A regression from 1.43
Support local $\ = "somestring" (#318), also a regression from 1.43
Fix special 5.6. associations for $, $\ as they are no SVs there (#306)
Fix wrong caching of internal LEXWARN symbols (#322)
Warn on bad IO::Socket::SSL versions 1.956-1.983 which would SEGV if used as server (#317)
Support __DATA__ blocks in packages without printing wrong warnings (#310)
Support ->(maybe|next)::(method|can) mro method calls (#324, #301)
Support mro c3 (#316)
Support @- @LAST_MATCH_START, the array of the last submatches (#281, #220, #295)
Workaround a 5.18 de-optimization which disabled ++PL_sv_objcount on bless
by adding it if a DESTROY method exists. This will fail since 5.18 if DESTROY
is added at run-time. (#208, #197)
Add new init0 section to initialize -fppaddr ops before init, not afterwards. (#330)
Needed to call utf8::SWASHINIT at init with CALLREGCOMP and /i
Avoid duplicate PL_utf8_tofold ToCf swash initialization with 5.18 (#330)
* CC (1.15): Skip saving non-existing methods analog to B::C 1.43_06 (CC test 50)
Fix failing CopFILE_free and CopSTASH_free in END blocks with threads (#296, CC test 48)
Handle duplicate function names, like multiple END blocks or anon functions (#297)
Replace each %hash with foreach sort keys (#307)
* Bytecode (1.15): Replace each %hash with foreach keys (#307)
* Bblock (1.04): Do not pollute B::CC (find_leaders) with B::Concise
1.45 2014-02-11 rurban
* t/issue281.t: fix wrong test ($[ vs $])
1.44 2014-02-10 rurban
added adjusted CORE testsuite for 5.14. passes for 5.14.4-nt, not for 5.18.2-nt
(cPanel code_monkeys)
still missing in general are:
- global destruction (use lexicals!),
- attribute handlers,
- compile-time perlio layers
* C: Fix -O3 with ~ and ~~ formatstrings (#277)
Fix SvLEN and PV ptr for empty shared hash keys (#272)
Use the natural @dl_modules boot order, just put attributes to the front (#259)
Store cop_hints to support lexical numeric hints pragmas,
esp. use bytes and use open attributes
(fixes #81 #152 #174 #176 #180 #185 #224 #235 #238 #241 #245 #253 #255 #256 #277)
Skip saving defined(&cv) subs (#96)
Fix format STDOUT/STDERR (#238, #239, #277, #284, #285, #283)
Save now empty bodyless subs if they exist, for signal handlers,
prototypes declarations and cvrefs (#251, #159, #235, #246, #270, #271, #275, #279)
Fix PVMG PV overwriting the RV, overload sub (#273)
* CC (1.14): Skip saving non-existing methods analog to B::C 1.43_06 (CC test 50)
* perlcc (2.17): Accept -A for -DALLOW_PERL_OPTIONS
* t/testc.sh: Accept -A for -DALLOW_PERL_OPTIONS
1.43 2014-01-15 rurban
new 5.16, and partial 5.18, 5.20 and windows support.
Many more fixes and testcases.
Bytecode is still broken on 5.18, esp non-threaded. 5.20 looks pretty good though.
* C: Fix -u<module> without file extension.
With hek do not drop FAKE, only for const pv.
Fixed $$ ($PID) to be dynamic, issue 108. Thanks to flexvault for reporting this.
Fixed double precision to 16 digits. The nbody shootout test passes now.
Fixed refcounts of *ENV, issue 111.
Fixed wrong boot_ arg for the xs version check with --staticxs
Better fix for overwriting of @ARGV <O3, test c_argv:1 and c_allow_opts.t:1
Fixed refinition warnings of XS CONSTSUB. Ignore them as they are
loaded later.
Add $dlmodule::VERSION as 2nd arg to boot_$dlmodule resp. $dlmodule::bootstrap
to get rid of the pesky Invalid version format error in xs_version_bootcheck
if the $VERSION had already version magic attached. (no testcase, hard to repro)
perl 5.16.0 shipped a slightly broken B without cop_stashlen. As workaround
we do not support utf8 stashes nor null bytes in stashes on 5.16.0 threaded.
Fixed Null COP storage for 5.16 and 5.17
Fix amagic_generation which was removed with 5.17
Fix ALLOW_PERL_OPTIONS: add -e behind all processed options, not before
Fixed HvAUX init. Previously only with magic hashes, now on all OOK hashes (OOK+SHAREKEYS)
e.g. %warning::Bits
Added support for new PADLIST type since 5.17.4 (with xpadl_id) and 5.17.6 (without)
$^X returns now the real executable name, not just perl or perl.exe
Adjust COW string lengths to hold the COW_REFCNT byte at the end
Fix unicode string CUR and LEN (#142, #162)
Fix unicode hashkeys (HEK) mostly (#200), but not yet testc.sh 2001 (compile-time
stored utf8 heks)
Save HvTOTALKEYS at run-time as max added to the number of run-time added keys (#178)
Fix UV uvuformat on 32bit -Duse64bitint (#145)
op_pmreplstart is now run-time initialized >= 5.10 (#160)
Fixed special usage of magic vars @-,@+,%+,$-,$! (#90)
Fixed wrong REGEXP length for compiled QR data (#143)
Fixed wrong -O3 comparison of strings to numbers (#138,#141, branch new-cog), unified
PV handling, disabled -fcog replaced by -O3 -fconst-strings
Fixed NUL-byte handling in strings by using B::SV->PV instead of PVX (#237)
Storing now bareword filehandles (#148,#149)
Do not create GvGP at all. This also removes the need for newGP() (Windows/AIX)
use Config does not include B anymore, Internals::V now in __ANON__ package
Fixed handling of empty - not successfully autoloaded - CVs (#235, #159, ...)
by storing the SV and XPVCV but ignoring it and defer to run-time get_cv()
Unify handling of static and dynamic strings, use ptr_undef throughout instead
of NULL vs &PL_sv_undef (save_pv_or_rv => savesym, cur, len, pv, static)
Restore $^H, the numeric value of global warnings flags (#193,#207,#243)
Restore more missing globals: ${^UNICODE} ${^UTF8LOCALE} $; $\ $, $/ $" $|
$^A $^L $: $^H $^R $% $- $= if set in a BEGIN block or via cmdline (#256,#231)
Restore even deprecated vars $] $# $* (#171)
Fix HvAUX memory corruption, accessing HvAUX beyond HE fields, esp. with threaded perls.
Initialize now HvAUX backref and mro_meta fields.
Do not boot internal core XS packages twice, refer to internal XS functions dynamically
and do not link to them. Fixes fix Windows/AIX with strict linking.
Defer dynaloaded GvCV initialization after dl_init()
Fixed storage of non-special lexical warnings (#258 warnsize-i258)
Fixed storage of shared heks, support IsCOW_hek (i.e. method names)
Support new ReANY since 5.17.6 (#258 reg_temp_copy)
Support stash symtab magic for PMf_ONCE m?? and reset (#188)
Fixed POSIX overflow constants warnings for DBL_MAX,DBL_MIN,LONG_MIN,LONG_MAX (#262)
Support user-defined static_ext modules, which are already booted from core.
Add more DynaLoader::dl_debug output (set PERL_DL_DEBUG=1)
Enforce -ffold on an ucfirst op or $INC{'unicore/To/Title.pl'} (#242)
Mark empty but overloaded packages (#172). This is merely workaround for reduced testcases.
Set PadnamelistMAXNAMED(comppad_name) needed since 5.19.3 for eval string (#268)
Add new option -fno-delete-pkg to control if to ignore compiler-only dependent packages,
compiler-independent packages are now always saved. Save those package names at Makefile.PL
in @B::C::Flags::deps.
* perlcc (2.14): new option --dryrun, -v5 does not -Dsp,-v
new option -f passthru to C and CC
* perlcc (2.15): fixed default --spawn: use waitpid which was broken for parallel builds.
fixed warning Prototype mismatch: sub main::gettimeofday: none vs () at perlcc line 261
improved --version
added --perlopts to handle -DALLOW_PERL_OPTIONS
* perlcc (2.16): changed output name rules:
Without given output file name we use the name of the input file (in the subdir)
or with -e a.out resp. a.exe and a randomized intermediate C filename.
If the input file is an absolute path on non-windows systems use the basename.
* CC (1.13): Use the B::C integer and double precision logic (ivx, nvx).
Fixed double precision to 16 digits. The nbody shootout is now 2x faster than perl.
Added optimizations: -fno-magic, -fno-autovivify, -faelem
Detect "no autovivification;" pragma.
New -fno-taint, -fomit_taint is deprecated
Fix amagic_generation which was removed with 5.17
Use new perl6 type names: int, num, str. double and string are deprecated.
* Bytecode (1.14): fixed require and op_first, issue 97
Fixed regex_pad offset in threaded perls >= 5.11, issue 68.
New type B::PAD isa B::AV (PADLIST for 5.17.5),
New bytecodes newpadlx, padl_name, padl_sym (PADLIST for 5.17.5)
Fixed CvGV_set causing Attempt to free unreferenced scalar in push_begin (42,43)
Fixed -Do (peek ops)
Renamed option -f to -F for files.
Fixed READONLY magic and restricted hashes, issue 98
Support IsCOW xpvshared strings (>=5.18)
Fixed wrong xpvshared logic (5.10-5.17), issue 138
Optimized away 0=default op and gp flags and pointers
Optimized away default sv_refcnt=1 (GV not), and skip redundant bytecodes
Set PadnamelistMAXNAMED(comppad_name) needed since 5.19.3 for eval string (#268)
* ByteLoader (0.10): set sv_refcnt to 1 in newsv to skip most defaults
* Assembler (1.11): allow "newpadlx 0"
* Disassembler (1.12): use B::Concise op_flags and private_flags
* Stash (1.03): fix compilation for 5.8.8 and below: gv_fetchsv missing
* t/perldoc.t: perlcc fails with 5.8 because Cwd disturbs the
fragile method package finder for File::Spec. Use cc_harness.
1.42 2012-02-01 rurban
stable up to 5.14
* C: Improved finding methods in parent classes (Warning: method not found),
Save @ISA of those child classes.
Re-implement and simplify my_share_hek, now in terms of share_hek.
PL_strtab hack gone. This fixed the shared_hek destruction
problems.
Turn off CvDYNFILE needed since 5.15.4, use static CvFILE. (DateTime)
* perlcc: perlcc -stash does not hang anymore, renabled t/e_perlcc.t -stash section
* Stash (1.02): '-xs' does not load B anymore.
Now in seperate directory Stash
* Stash/Stash.xs: new with CvIsXSUB() to get rid of B, polluting the stashes.
* Stash/Makefile.PL: new
* Bytecode (1.13): support shared hek (42,43)
* ByteLoader (0.09), ByteLoader/bytecode.h, bytecode.pl: support shared hek (42,43),
added xpvshared
* t/issue95.t: added, isa methods not found
1.41 2012-01-30 rurban
* t/modules.t: Fixed syntax error
1.40 2012-01-30 rurban
static PerlIO Layers (e.g. __DATA__)
* C: fixed deferred %INC, %INC is now stored and cleaned up at the end.
Improved __DATA__ handle as PerlIO layer, PerlIO::scalar now
loaded static (test 15).
Defer eval section after dl_init (e.g. for PerlIO::scalar) (test 15)
Check for XSLoader now at the very end (test 27)
Fixed MUTABLE_CV for 5.10.0
Improved 5.8 method lifting for not-found methods a bit (5.8. perldoc.t)
* Bytecode: fixed test 22
Cannot locate object method "fileno" via package "IO::File"
* Makefile.PL: readonly $lib fix for MSWin32
* t/testc.sh: silence ulimit -m warnings + errors
1.39 2012-01-25 rurban
* C: improved eval AUTOLOAD, avoid some POSIX and Storable .al compile-time
side-effects such as creating files and dirs. First try loading
.al files, only then eval the dangerous AUTOLOAD.
* Makefile.PL: typo in -fav-init2 independent_comalloc detection.
Added LICENSE key, META and the files were not enough to please pause.
1.38 2012-01-24 rurban
* t/issue93.t: fixed syntax error
1.37 2012-01-24 rurban
Enabled copy-on-grow strings >= 5.10.
Save %main:: stashes w/o -fno-stash. Switched to git, added github mirror
Fixed hashes and share_hek, when mixed compiled and load-time for >5.8.
Stability: This release is the first which passes almost all tests and
is used in production with --staticxs -O3, but copy-on-grow hek's and
further logical improvements (deferred %INC cleanup) fail so far, so it's
still considered unstable.
* C: save stashes optionally with -fstash, but only values which are also
stashes to avoid a second walker besides savecv, and thus do not
walk into unneeded territory; B::STASHGV (46, issue 79)
New -fno-stash with -O2 to omit stashes which are rarely needed.
Fixed vulnerability to oCERT-2011-003 style DOS attacks. Do not
pre-compute hashes for hv_store(), use fresh random hash seed.
But shared_hek() has a limited API which does not allow this.
UNIVERSAL methods were not correctly stored, which led to method not
found at run-time errors.
New debug option -Dr "runtime" to add debug output to the .c code (as in CC)
New option -Upackage to skip package, mark_skip()
Re-implemented sharepvn via share_hek() and SvLEN=0 (issue 71),
Special FAKE+READONLY handling: LEN = 0 if pv_copy_on_grow or shared_hek.
Fixed !SvIsCOW_shared_hash + FAKE+READONLY conflict (test 13, issue 82).
Shared COW hashkeys which are optimized to static const strings
may not be marked as IsCOW (remove SVf_FAKE).
Re-implemented CORE share_hek with hash=0
Otherwise shared HEK's (post-5.10) come up with duplicate entries
for the same keys; one compile-time created and one run-time created with
a different random hash seed. E.g. observed by such warnings
"Prototype mismatch: sub bytes::length (_) vs (_)"
Added Internals::hv_clear_placeholders XS name exception, similar
to version:: (issue 83)
Store CV prototypes, issue 81 and issue 84 for empty () vs NULL prototypes.
Fixed 5.15.2 hang at hfree_next_entry (issue 78)
Defer writing of READONLY hash keys since 5.15 (issue 88)
Enable -fcog copy-on-grow with static strings for >= 5.10
Improved Errno vs. *main::! and Tie::Hash::NamedCapture vs
*main::+ / - logic (issue 90)
Updated version XS methods, default to noop for (...
Strip version object overload from XS packages (issue 91), XS needs plain PVs
Do not initialize CONSTSUB XS methods twice, boot does it for us
Pass proper cv to XS boot_$name, fixes boot of
Tie::Hash::NamedCapture (issue 86)
Fixed share_hek strings with -O1 (test 13)
Fixed bootstrapping the XS/core part of attributes,
thereby fixed threads tests 41-43.
Fixed most 5.8.[45] problems by ignoring aelemfast SPECIAL pads.
Fixed magic -> PTR when SV (>5.6)
Moved -fsave-sig-hash from -O3 to -O0. You need -fno-save-sig-hash
to disable it.
Removed -fno-stash from -O2, -fno-stash is now the default. You
need -fstash to add stashes. It caused not understood problems.
Fixed saving %SIG signals and more -2 SV magic (issue 92)
Adjusted IO logic, was erronously skipped (issue 59)
inc_cleanup: save only those %INC keys which are stored.
Save ext/mro only if the ext mro methods were loaded, and
utf8_heavy.pl similarly.
Force loading of -u arguments to avoid runtime loading.
Call boot_core_PerlIO, mro, UNIVERSAL and xsutils.
Warn on unsavable IO objects from BEGIN blocks.
Fixed >&STD... handles (issue 93),
i.e. Test::Builder, Test::NoWarnings and parts of the core testing.
See http://blogs.perl.org/users/rurban/2012/01/what-to-avoid-in-begin-blocks.html
Add B::COP::stashflags workaround, >5.15.4 threaded.
Add fix for lost PL_regex_pad, 5.15 threaded.
* CC (1.12): allow overriding of -f<opt> B::C flags.
Skip internal packages with mark_skip (fixes cc_o2 tests).
Move -ftype-attr from -O2 to -O1.
Always define int, double and string package versions, parallel to types.
* Bytecode (1.12): detect Tie::Hash::NamedCapture (issue 90) -
works only partially.
Fixed >&STD... handles (issue 93)
Support -d debugging without -MOd
* Stackobj (1.00_03): add UL or L to overlong integers. doubles
not yet
* perlcc (2.13): added options -O[1-4], -u, -U, --Wc, --Wl, --version,
Keep cfile if output is empty.
Change -o output C file directory and name with -S or -c:
Before in current directory, now in -o directory.
Before with input name, now with -o name.
Add -rpath for all --staticxs libs if supported.
Changed verbosity levels and output wording.
* bytecode.pl: added xio_ifp
* bytecode.h: added BSET_xio_ifp
* META.yml: remove from MANIFEST and repo, let make dist
autogenerate it.
* t/test.pl, t/TESTS: changed test 46 (Exporter:: and other stashes in
%main:: stash) to cover issue 79.
use Exporter; print q(ok) if %main::Exporter::
* t/test.pl: use general comments with optional TODO. Before only TODO
* t/perldoc.t: fixed for 5.8 with manually adding -uFile::Spec -uIO::Handle
* t/issue59.t, t/issue81.t, t/issue90.t, t/issue93.t: added
1.36 2011-10-19 rurban
lexical warnings and more 5.15 fixes
* C: new B::LEXWARN, fixed representation of lexical warnings which was broken
since 8e01d9a6 5.8.9/5.9.4 (issue 76)
Added section free, use it for lexical cop_warnings (const STRLEN*)
(TODO: static PV, HEK + COPs)
Workaround for utf8::SWASHNEW adjusted to 5.15
PAD* sv flags changed in 5.15.3
XSLoader API changed (HEK in caller i.e.): had to add a
XSLoader::load_file for [perl #101336] >= 5.15.3,
also use mXPUSH for sp instead of 1-arg targ
Set SvLEN=0 for static PVs to skip sv_clear (problem with 13)
Handle overlarge negative 64bit ints, double not yet.
Move -fwarn-sv from -O2 to -O1
Move -fsave-data from -O3 to -O2
Compile *main::DATA IO handle automatically (__DATA__), just package __DATA__
handles need -fsave-data (15)
Die with 5.8.4/5.8.5 threaded with split->pushre on invalid
pmreplroot. This is a core B::walkoptree bug and can not be
easily fixed.
Tests 15 and 103 do not hang anymore, 15 and 29 pass mostly
* C.xs: added -Dts support to runloop to check internal compiler
problems
* t/issue76.t: added
1.35 2011-10-02 rurban
tested from 5.6.2 until 5.15.3
* C: allow debugging without -MOd=C
improve package_pv detection for methods, two more tests.
detect previously missing packages within the main sourcefile
(50,35,72,73)
-O1 now includes also -fav-init2 and -fppaddr, which are all
stable. -O2 string handling (-fro-inc) still unstable.
added B::REGEXP::save for \$qr (Encode::Alias), issue 71. Still
not fully fixed. Encode::Alias unusable.
fixed empty HV assertions since 5.15, new HV default keys = 7
(3,4,36), but still unsolved 5.15 problem with endless loop in
Perl_hfree_next_entry hv.c:1716 (3,4,36)
workaround 5.15.2 inability to walk with from %INC deleted
stashes, also workaround adding already deleted DynaLoader.
CORE patch [perl #100138] for new 5.15.2-5.15.3 DynaLoader XS_INTERNAL
problem required (16,29,44,45). Fixed with 5.15.4
0001-Export-DynaLoader-symbols-from-libperl-again.patch
use proper PL_envgv for *main::ENV which is inited in perl_parse
(fixes e.g. CGI apps using run-time values, not compile-time)
* CC (1.11): allow debugging without -MOd=CC
Try to jump from last to unknown label, put labels also onto cxstack.
Fixed cc_last.t test 4, jump out of anonsub, but not across C
functions yet (this is disallowed in C, need to split).
* C.xs: added B::REGEXP::EXTFLAGS (missing from B).
* perlcc (2.12): omit also inc_version_list dirs from perlcc as
leaving them in could lead to unexpected crashes
* t/cc_last.t: Fixed test 2. This works compiled and uncompiled,
but the returned errcode is not compared. Skip if so.
* t/TESTS: add all 5 possible method/sub calls to test 35 =>01234.
See http://blogs.perl.org/users/rurban/2011/06/how-perl-calls-subs-and-methods.html
* t/stash.t: fixed 5.8.8 stashes (overload, threads, ...)
* t/issue71.t: added, but not fixed yet
* ramblings/*.patches: CORE patches added and recommended in README.
* ramblings/blogs-debugging-article[1-4].pod: added to MANIFEST.
* perlcompile.pod, perloptree.pod: improved.
* ramblings/yapceu_2010.pod: added.
1.34 2011-06-12 rurban
* Makefile.PL: fixed make install < 5.13.7
* issue24.t, test.pl, bytecode.t: TODO more failing tests from cpantesters
1.33 2011-06-12 rurban
* Makefile.PL: fixed make install > 5.13.7
* test.pl, bytecode.t: TODO more failing tests from cpantesters
* stash.t: fixed freebsd test 4
1.32 2011-06-10 rurban
more darwin and package scan fixes, Bytecode -i includeall.
all top100 modules can be now be compiled on most perls
* Bytecode.pm (1.11): fixed const xsub xsubany ptr >=5.10 <5.14 (27).
save_cq => save_begin+save_init_end: moved push_begin upfront for the same
init order (44).
add CVf_CVGV_RC flag (refcount) on CV with no backref magic, e.g. END (48)
added option -i includeall which adds no BEGIN require ops, but adds
all included symbols.
better BEGIN block @INC manipulation detection (33)
print newsvx flags if -S or Comment
print hex IV < 8 as int with -S
* C.pm (1.32): check PERL_TRACK_MEMPOOL for AV malloc (25).
fixed init chunk splitting for CV and -fav-init2, if >10000 lines.
fixed overlarge AvFILL=3 of endav. Undefined subroutine &main:: (48)
fixed shared GPs on typeglob assignment. do not gp_free shared GPs,
just set them once.
added try_isa before try_autoload. use mro::get_linear_isa instead of @ISA
mark_package force: detect and fix already discarded packages which
are brought back in directly or by someone else's @ISA. (e.g. Sub::Name
needs DynaLoader::dl_load_flags)
improved method_named package search, still not exact though (LWP)
add manually compiled packages to %INC to avoid a runtime require
special case Config detection, AUTOLOAD = launcher, included by Dynaloader
added -Ds to print all sub names
do not compile any B::C subs, esp. our %SIG handler
add our own Internals::V as it is not exported by libperl
inc the REFCNT of stashes (Path::Class)
* Assembler.pm (1.10): archflag bit 2 for MULTPLICITY
* Disassembler.pm (1.10): archflag bit 2 for MULTPLICITY
use longsize resp. ivsize from .plc not from perl
print newsvx flags
print hex IV < 8 as int
* ByteLoader (0.08): support archflag 2 MULTPLICITY,
[perl #86186] force binary IO for __DATA__ handles, which changed
in core with 5.14
* bytecode.pl: support m for MULTIPLICITY (yet unused)
* Makefile.PL: move cc_runtime.h installation to make install step
(sudo). Tested ok on darwin gcc, does use -I. internally.
* cc_runtime.h: renamed from cc_runtime514.h
* cc_harness: fixed darwin coredir
* perlcc (2.11): fixed darwin coredir.
omit darwin linking to dynamic .bundle - generate .a via libtool
* TESTS: removed END block from test 28, only test run-time require (fixed 28)
added 48 for previous test 28 END block lexvar del_backref failure
added 49 for m//i blowup and -fno-fold testing
added 50 empty @ISA issue 64
* assembler.t: fix for hex->int IV roundtrips
1.31 2011-03-21 rurban
multiplicity support and more darwin fixes
* C.pm (1.31): replace $ITHREADS by $MULTI (darwin),
my_perl_destruct: disable CopFILE_set NULL and CopSTASHPV_set NULL
on $MULTI, PL_sv_yes|no special on MULTI
* CC.pm (1.10): replace $ITHREADS by $MULTI (darwin)
* t/test.pl, issue34.t, issue35.t: -ofile => -o file (darwin cc)
* cc_harness: -Wl,--warn-once invalid on darwin gcc
* log.modules: encode MULTI without threads as -m
1.30 2011-03-06 rurban
dbmopen, cc_runtime.h, perlcc, --staticxs
* C.pm (1.30): support dbmopen, fixes issue 24.
fixed xpvmgsect (was xpvsect) !-fcog < 5.10 (YAML, test 41,42,43).
fixed m//i since 5.13.10: swash_init("utf8::Cased").
fixed method_common package_pv detection in most cases.
still two cases open, without tests. PADOP GV NULL 0x20000, GV RV 0x40802.
fixed method_named with >= 1 arg, checking for const PV package_pv (test 35)
support GvCV_set and GvGP_set introduced with 5.13.10. [RT#65630]
implement -staticxs for perlcc --staticxs
either link to found static libs or link to shared lib with
fixed LD_LIBRARY_PATH/PATH (nyi in C, just perlcc).
optimize -fav-init use_av_undef_speedup &PL_sv_undef initialization (again).
reserved -mPackagename (NYI)
added verbose(),module() for CC.
implement outline of -m module, name detection missing.
added -fno-fold recommended since 5.13.9 to omit the huge utf8 tables.
added -fno-warnings, recommended since 5.13.5.
added special workaround for Scalar::Util bootstrapping which did not detect
List::Util, since it is not in its ISA, only required. need a better scan.
fixes test 44.
init XS within Od.
prevent autoloaded STDERR printing on non-debugging perls (e.g. "Storablelogcarp").
* CC.pm (1.09): support dbmopen, issue 24 failing due to XSLoader (test 45)
no pp_name shortnames in embed.h since 5.13.9, use Perl_pp_name(aTHX).
implemented dorassign, fixes issue 45.
fixed XSLoader detection, moved use_xsloader to save_unused_subs()
skip duplicate cc(pp_sub_*) functions (freebsd 45)
added missing END block code (test 28).
added B::C::verbose on -v
fixed B::C::verbose crash by adding B::FAKEOP::fake_ppaddr.
fixed compile_stats
renamed -n module_name to init_name.
check and set -f options from C also.
implement outline of -m module, name detection missing.
added option -fno-name-magic.
added experimental option -ftype-attr.
added TYPES pod section.
added pod for some funcs: load_pad, cc_queue, label.
option -strict only used with DEBUGGING perls.
be less chatty with compiling bblocks.
light layout cleanup of generated code sections.
added Vishal Bhatia <vishal at deja.com> to AUTHORS.
* Bytecode.pm (1.10): support dbmopen, fixes issue 24.
re-added -m, compile as module to .pmc
store AV elems for non-tied AV->MAGICAL >5.6 (i.e. @ISA, ...) (new test 47).
-DA: limcheck PV.
verbosity: added nice1 (condense elements).
name SV indices also.
* Byteloader: support GvCV_set introduced with 5.13.10 [RT#65630]
* perlcc (2.10): fixed -B -r, -B output and run bytecode.
implement --staticxs to workaround DynaLoader problems. use it in t/modules.t
add sopath to PATH on -r. Without -r the sofile should be symlinked so that the
process can find it.
implement --shared and --static, --shared fails on strawberry due to core problems.
implemented getopts bundling: -ScO -e'bla' -v4 -oa is now possible.
disabled some old single-minus switches: -testsuite, -time, -log, ...
requires now double-minus --testsuite, --time, --log, ...
kept legacy handling only for -Wb=.. and -stash.
fixed -B -e'oneliner'.
fixed -B @ARGV, no duplication of @ARGV, c_argv.t 3.
re-added -m|--sharedlib for Bytecode and C (NYI).
print C and CC filtered compiler output.
-v5 adds -v to the compiler options.
-v6 adds -Dfull to the compiler options.
support -Wb= for -B.
fixed -Br for 5.6: add -MByteLoader.
* Stash (1.01): add option xs to print xs modules with -x prefix.
usable via -MO=Stash also.
new option -D for debugging.
omit PerlIO::Layer.
added pod
* Disassembler (1.08): 5.6: fix ldop comments.
add op_type names, add @svnames, add indices, add ldspecsvx type.
* Assembler (0.10): print more flags (as hex) and indices.
better limcheck diagnostics (op,sv,pv).
* cc_runtime514.h: removed from CORE with 5.13.9. [RT#65628]
If so, cp it back. You might need sudo cp. (unchecked)
Renamed not to pollute tests with older perls.
Expanded PERL_MAGIC_taint 't' for 5.6.
* C.xs: added experimental method_cv for the hash-only case, untested.
* t/modules.t: try --staticxs first, try crosscheck without perlcc.
* t/issue24.t: added.
* t/issue45.t: fixed and added 3 more tests.
* t/e_perlcc.t: added to test the new option handling.
* t/testc.sh, t/TESTS: fixed wrong test 39 for 5.8, added 47, 104 (reset), 105 (type-attr).
* t/testplc.sh: added 47.
* t/stash.t: rewritten. allow 5.6, use Test::More, more stable
* t/todomod.pl: added. interactive tool to check log.modules reports and fix TODOs
* status_upd: recommend and fix ./status_upd -f -q -d
-q -d prints only the actual tests
* Changes: added more missing history ca. 1997-2000.
1.29 2011-01-08 rurban
improve dl_init
* C.pm (1.29):
use DynaLoader::bootstrap_inherit, fixes dl_init when the module has no ISA
DynaLoader. eg. Test::Harness with Time::HiRes
improved CopFILE detection as context for dl_init (test 29)
fix cop_warning cast warnings
print debug messages for beginav and endav
protect against undefined strings, and % in regex
protect against undefined gvcv->GV->EGV (Spiffy)
latefree r-magic regex strings >5.10, fixes a couple of modules
* t/modules.pm: subset use 10 not 11 modules. run tests with -O1
less TODO tests, esp for 5.13
* t/modules.t: fixed binary detection on non-windows (a.out)
* ByteLoader: fixed -Dt for GV
1.28 2010-12-31 rurban
fixed XSLoader, pmflags, CV and most named methods for 5.14
* Bytecode.pm (1.09):
fixed cv_gv and push_begin del_backref problem (27, ...) > 5.13
special cased op_pmflags: U32 needed since 5.13 (PMf_BASE_SHIFT)
Fixes op_pmflags since 5.12 ((3..4,27..29,31,33), esp. pp_subst
start for Windows 5.12 cop_store_label workaround: Labels disabled for now.
Failed to include CORE hv.c: too large.
mingw 5.12 compiles now again and passes all but 21,33,42..44
Fixed entrytry with Bytecode >= 5.12:
ck_eval upgrades the UNOP entertry to a LOGOP, but B gets us just a
B::OP (BASEOP)
add BEGIN { push|unshift @INC } parts of blocks (test 33)
* Byteloader (0.07):
Support new CvSTASH_set.
Basic op_pmflags U16/U32 support. No version conversion yet.
BSET_gv_fetchpvn_flags: new for empty constant prototypes (fails on 64bit only)
* C.pm: add -fno-destruct (with -O3) with a minimal perl_destruct,
(re-enables -fcog >= 5.10),
add -fro-inc (with -O2) readonly INC and curpad strings,
add -fconst-strings (with -O3) const readonly strings
and const more other fixed strings,
do not boot static core packages (utf8, re, ...),
fixed -O2/-fav-init2: store at wrong avchunk index,
initialise CopFILE and CopSTASHPV static with ITHREADS
improved 5.13 method_named package detection (first and pmreplroot),
in endav still failing (Test::Deep)
fixed overwritten CvSTART for 5.13.x
do not cog/const nullify vars in END blocks
disallow CONST->op_first with CONST_BARE. fixes ExtUtils::Install
better XSLoader::load detection: aliases and defined in END blocks
save XSLoader::load GV (test 46 and modules)
bootstrap also static core packages
add ptr_undef instead of NULL initialization non-threaded
fix many cast warnings
fix AV fill +1
required to run C.xs runloop to find PMOPs for QRs
Fixed entrytry >= 5.12 as in Bytecode:
ck_eval upgrades the UNOP entertry to a LOGOP,
but B gets us just a B::OP (BASEOP)
Fixed XSLoader, loading dynamic modules works now (test 45),
had to set the .pm path in the calling context.
* CC.pm (1.08): many fixes
use new B::C optimizations, new -DF
fixed issue 31 also run C.xs runloop to find PMOPs for QRs,
fixed issue 35 same variable name in different scope (Heinz Knutzen),
fixed issue 36 panic: leaveloop, no cxstack (Heinz Knutzen)
fixed issue 37 orassign (Heinz Knutzen)
fixed issue 38 on and/or return value not just no/yes (Heinz Knutzen)
fixed issue 39 Bizarre copy of ARRAY in leavesub (Heinz Knutzen)
fixed issue 42 remove enter/leave from %no_stack (Heinz Knutzen)
fixed issue 44 aelemfast missing for lexicals (Heinz Knutzen)
fixed issue 47 anonsub in while (Heinz Knutzen)
fixed issue 48 wrong truth value for array assignment in boolean
context (Heinz Knutzen)
fixed issue 49 Can't "last" outside a loop block (Heinz Knutzen)
fixed issue 51 errors on nested if statement with test on multiple variables
(Heinz Knutzen)
fixed issue 52 errors on variable with numeric value used in second
expression of 'and' (Heinz Knutzen)
add labels for alternate ops (e.g. run-time jumps). Not sure yet if the
full alternate subtrees are saved. fixes test 32 (push_label, pop_label),
issue 46. resolve duplicate such labels.
honor $ENV{'PERLMODS'} (by cPanel)
* Disassembler.pm (1.07): read over #! + use Bytecode .plc header.
support pmflags
* Assembler.pm (1.09): support pmflags
* perlcc (2.09): added cc time for --time.
same INC as in caller - omit duplicates
* cc_harness: improve -Bstatic
* t/testc.sh: simplified. Fixed various wrong tests
* t/testm.sh: fixed UAC issues on Win7
* t/test.pl: new faster test functions for the new issues
* t/issue27.t: added (Reported by alexchorny, Apr 25, 2010)
* t/issue29.t: added (Reported by Heinz Knutzen)
* t/issue31.t: added (Reported by Heinz Knutzen)
* t/issue34.t: added (Reported by Heinz Knutzen)
* t/issue35.t: added (Reported and partially fixed by Heinz Knutzen)
* t/issue36.t: added (Reported and fixed by Heinz Knutzen)
* t/issue37.t: added (Reported and fixed by Heinz Knutzen)
* t/issue38.t: added (Reported and fixed by Heinz Knutzen)
* t/issue39.t: added (Reported and fixed by Heinz Knutzen)
* t/issue42.t: added (Reported and fixed by Heinz Knutzen)
* t/issue44.t: added (Reported and fixed by Heinz Knutzen)
* t/issue45.t: added (Reported by Heinz Knutzen), dorassign missing
* t/issue46.t: added (Reported by Heinz Knutzen), fixed by r610
* t/issue47.t: added (Reported and fixed by Heinz Knutzen)
* t/issue48.t: added (Reported and fixed by Heinz Knutzen)
* t/issue49.t: added (Reported and fixed by Heinz Knutzen)
* t/issue50.t: added (Reported by Heinz Knutzen)
* t/issue51.t: added (Reported and fixed by Heinz Knutzen)
* t/issue52.t: added (Reported and fixed by Heinz Knutzen)
* t/issue54.t: added and fixed
1.27 2010-07-30 rurban
Fixed 1.26 CV regressions for 5.8 and 5.10
* C.pm: add -DF print stringified sv and op flags.
fixed xpvio off-by-one error for 5.10 (test 29)
rewrote XPVCV forwards for >=5.10: (fails before)
1. only the SV is now forwarded, not the struct.
2. no duplicate SVs, the CV forward is just a link to one SV (#define CVIX)
removed wrong CVf_ANON check causing CV failures (test 9,10,12). 5.13.3 still failing
fixed CV svsect for 5.8 and 5.10
* Bytecode.pm (1.08): removed wrong CVf_ANON check causing CV failures (test 9,10,12,...)
* perlcc (2.08): added --time, support vprint -1
* t/test.pl, t/bytecode.t: double check failing tests if it works uncompiled
* t/testc.sh: add -D- for no gcc warnings
added -f for testing single optimizations
* t/testm.sh: add -D<args> flags for compiler debugging and link with -g
added -T for perlcc --time
* t/c_argv.t, script/perlcc: properly quote spaces in path of perl and `cwd`/blib
* t/modules.t: add svn rev to B::C::Version
1.26 2010-07-26 rurban
Start of 5.14 support, CVs broken.
* C.pm: improve -fav-init performance if not usemymalloc: no
calloc, just direct libc malloc.
-O2 uses -fav-init2, i.e. experimental independent_comalloc(),
-O1 uses now -fav-init (Nick Koston + Reini Urban)
Mark a package which is autoloaded from XS.
Fix autoloading of constants AKA const xsubcv via AUTOLOAD [test 27] (Nick Koston)
5.13.3 support (xpv STASH, MAGIC, ... reorganization), but CVs still failing
Check CVf_ANON when setting cv->GV. Fixes anon subs.
Do not call REGEXP IVX and NVX methods (since 5.11)
* bytecode.pl: CvGv_set since 5.13.3
Fixed byteorder 0x stripping causing wrong bget_swab on 64bit
* B::C::Flags: new. added by Makefile.PL probing independent_comalloc()
* CC.pm: the default -fno-slow-signals adds PERL_ASYNC_CHECK at
the same ops as with 5.13, a major improvement.
* Bytecode.pm (1.07), bytecode.pl: xpvav.xiv_u was removed with 5.13.2 but not
used anymore. read and set av_flags only until 5.12, compile only before 5.10
Check CVv_ANON when setting cv->GV. Fixes anon subs
* bytecode.h: do not PM_SETRE with empty arg. fails since 5.13 debugging
* t/testc.sh, t/TESTS: fail test 44, change threads tests 41-43 to TODO
* perlcc (2.07), cc_harness, t/test.pl: debian specific fixes for their broken
-lperl linker args for ExtUtils::Embed::ldopts.
* t/test.pl: Have complete c/cc TODO/SKIP status in one place.
Fixed killing hangling tests with IPC::Run.
Fixed new cc TODO.
Run CC test 18 even not as AUTHOR
* t/c*.t: make test 27 easier to debug (Nick Koston)
Added test 45 (Nick Koston)
1.25 2010-04-11 rurban
* C.pm: remove global my_perl, pass it properly around.
* t/c*.t: added test 40 \000 byte in PV failing on ori 5.6.2 (Nick Koston),
added more magic tests for shared vars - n,p,P magic: 39-43. 41 nyi
* t/modules.t: fix -t (run module tests), print header only on full test
* t/modules.pm: do not test core or deprecated modules, esp. do not download perl
Added is_subset.
* t/testm.sh: implement -k
* t/testc.sh, t/TESTS: add 44 for weaken import and magic_killbackrefs #72922
1.24 2010-02-15 rurban
* C.pm: Fixed off by one array access in magic array init (34)
- valgrind is your friend.
Fixed aelemfast padsv access crash (Text::Tabs on freebsd7)
Fixed wrong stash names in gv_stashpv analog to gv_fetchpv
(Text-Tabs+Wrap t/dandv.t)
* t/testm.sh: implement -s without log file
* t/modules.pm: export skip_modules, silence module scanning STDERR
1.23 2010-02-14 rurban
* t/testm.sh: chmod +x
* t/modules.t: fix $have_IPC_Run
1.22 2010-02-14 rurban
Tested with ActivePerl 5.10.0
* bytecode.h: unshare_hek exception for WIN32, not only MINGW
* t/modules.t: Added list of to be skipped modules, which cycle endlessly.
Improve todo message.
* t/testm.sh: add option -s "install skipped modules"
* Makefile.PL, cc_harness, perlcc: fix ldflags for ActivePerl 5.10.0
for MSVC6 (ccversion 12) -opt:ref,icf leads to invalid outfile "pt:ref,icf"
* c+cc test 29 fails on 5.10 randomly. Make it a TODO
1.21 2010-02-14 rurban
* C.pm: Add missing Nullop for CC on 5.11.4 (cc 12).
Test 38 passes
1.20 2010-02-14 rurban
* C.pm: #define PERL_CORE broke Null macros on 5.11.4. Define those.
1.19 2010-02-13 rurban
Almost fixed evaltry (12) with proper cop_seq setup, just solaris and debian
fail sometimes. Tested ok with strawberry 5.10.1
* C.pm: fixed evaltry (test 12) mostly, NVX was shared with 2 xpad_cop_seq ints
which accidently just worked before 1.17 with %s, but not after changing
to the %g representation.
Ignore unused random (overlong) xpviv IV, causing
"warning: this decimal constant is unsigned only in ISO C90"
Defer rv => cv to run-time < 5.10. Fixes ExtUtils::Install
Defer xpvmg->pv init to run-time < 5.10. Fixes MooseX::Types
Check for NV inf. Fixes compilation of DateTime
Defined PERL_CORE: Fxied Windows problems with setjmp undef'ed in XSUB.h
Added Win32 workaround for missing PL_do_undump
Stripped perl_ prefix from perl_get_cv, perl_call_pv, perl_call_method
* CC.pm (1.07): added -fslow-signals.
Disabled inlined pp_enter: Fails on empty cxstack.
Backed out inlining of enterloop: GIMME_V fails on freebsd7
Remove strawberry PerlProc_setjmp definition, #define PERL_CORE fixed that
* Bytecode.pm (1.06): same NVX => xpad_cop_seq fix as in C.pm, added 2 new
bytecodes (cop_seq_low, cop_seq_high), fixed tests 9,10,12. Passes
all tests > 5.6 now.
* bytecode.pl: added cop_seq_low (155), cop_seq_high (156) instead of xnv.
No conversion code for loading older bytecode needed, but fragile (double => 2 int)
* t/modules.t: refactored by Todd Rinaldo (toddr). 4*tests per module:
-s a, exitcode 0, ok, no warnings on stderr
* t/TESTS: added 38, failing on CC (Nick Koston).
added 39, failing everywhere.
* t/test.pl: refactor: Try to timeout on all tests (compiler and exec) if
IPC::Run is installed
* MANIFEST, *.pod: Moved conference pods to ramblings,
removed lib/B/Asmdata.pm again, added t/qr_loaded_module.pm
* t/testm.sh, t/modules.pm: added, -t not working yet though
* t/b.t: $[ => $]
1.18 2010-02-06 rurban
evaltry (test 12) still broken on most systems.
* C.pm, C.xs: fixed r-magic for 5.6 (ExtUtils::Install,
File::Temp, Template::Stash). need tests for run-time checks (33).
5.6 misses MG->pregcomp
* t/modules.t: check for a.out (test from toddr)
* Makefile.PL: create lib/B/Asmdata.pm beforehand (toddr)
* ByteLoader: fixed duplicate + missing bl_header for bsd and Darwin
(test from toddr)
* test 37 added - AV self-refs, analog to HVs (never failed, just to
be sure).
Removed -qq from 5.6 tests.
1.17 2010-02-03 rurban
Fixed now most module failures, non-threaded at least: bootstrap was in
the way. strawberry perl fixes. Fixed test 29, broke evaltry test 12
* C.pm: Added option -Dfull for all full debugging info, set all options.
Skip saving XS bootstrap. Fixes most XS module failures. i.e.
Digest::MD5, Data::Dumper, ..
Fail: Attribute::Handlers, B::Hooks::EndOfScope, YAML, MooseX::Types
4 for 5.8.9-nt from the top100, and 3 for 5.11.4-nt: Attribute::Handlers,
File::Temp, ExtUtils::Install.
Fixed AV self-references causing endless loops, analog to HVs (test 27)
Fixed Getopt::Long crash with AV D magic (regdata),
save_magic returns now the magic types as string.
Fixed unescaped savere (test ExtUtils::Install).
Fixed GvSTASH "initializer element is not constant" (Test::Harness)
Fixed xrv_list "initializer element is not constant" (Class::MOP < 5.10)
* CC.pm (1.06): Added missing unexported definitions for MSWin32,
>=5.10: newGP, vivify_ref, prepare_SV_for_RV (tests 18, 29).
PerlProc_setjmp, PerlProc_longjmp (tests 12, 32).
Fixed cxinc for 5.11.0
* t/bytecode.t: mixed up 32 with 33.
* t/modules.t, t/top100: new module test, see also the log.modules*
results in the svn repo.
* t/mymodules: added as reference, executed optionally.
* script/cc_harness: ldopts -std if no -B<opt>
1.16 2010-01-27 rurban
use Moose and benchmark still fails. IO even a regression.
* C.pm: Fixed braces warning for empty xpvcv startfield (Nullany),
register int for warn_sv loop
Fixed hv init for magic hashes, pre-allocate array and xpvhv_aux. (test 34)
Fixed method_named (CPAN #53889, test 35) for B::C, better fix than for B::CC.
Added -DH for HV debuggging.
Fixed HV self-references causing endless loops, i.e. with use Moose, test 36.
* CC.pm (1.05): added experimental -finline-ops. Always on. Can be removed if
it works okay for the new inlined versions.
Inlined pp_enter, pp_reset, pp_regcreset, pp_stringify.
Fixed pp_padsv DEREF for 5.10.
Documented test coverage for each op.
Fixed wrong $PERL511 definition.
Added Opcodes support (no_stack), on CPAN as 0.01.
added pp_pushmark pp_break pp_continue to no_stack handling.
Fixed method_named (CPAN #53889, test 35) assuming there is only const
before method_named. No counter examples found so far.
Changed new OBJ => OBJ->new
More curcop->write_back protection.
* Stackobj.pm (1.00_02): RV fix for 5.11.
* t/testc.sh, t/test.pl: added CC specific tests to increase CC op coverage
after 100.
Switched test 32->101 (arith) with 33->32 (evaltry catch die) as it fails
with 5.11.3.
Added new qr test 33 (Nick Koston), test 34 for magic hashes, test 35 for
broken methodcall syntax (method_named), test 36 for HV self-references.
* TESTS: added test 32 (evaltry catch die): Fails on 5.11.3 C/CC only,
<5.11.3 it works okay. Also tests 33, 34, 35, 36.
* ByteLoader/Makefile.PL: removed cygwin gcc specific stack-protector removal.
It was a bug of mine.
Added more AUTHOR gcc -ansi warnings, just -pedantic not.
* Makefile.PL: Added more AUTHOR gcc warnings, also -pedantic.
Added optional Opcodes PREREQ_PM
1.15 2010-01-03 rurban
* t/test.pl: Fixed MSWin32 tests again. Lost the 1.13 patches with 1.14.
1.14 2010-01-02 rurban
* CC.pm (1.04): Fixed entereval/leaveeval for 5.10. Fixes eval "", test 12.
The end of the string CALLRUNOPS needed to be nullified, as in 5.8.
5.11 CORE also needs a patch for this.
Optimized away SP save/restore for %no_stack ops. Need to read this
from opcode.pl eventually.
PERL_ASYNC_CHECK (pending signals) added after every basic blocks with -O1
and every loop.
* C.pm: More 5.10 union braces.
* Stackobj.pm (1.00_01): Added iv and nv to peek.
Added documentation.
* Bblock.pm (1.03_01): Print the argument (sub) to inspect.
Added missing documentation.
* t/testcore.t: improved. added -fail and a list of failing tests.
C now always with -O3.
* t/c_allow_opts.t: Added for -DALLOW_PERL_OPTIONS.
1.13 2009-12-31 rurban
Tested with MSVC (MSVC compiles test29 and bigger examples awfully slow!
7:30min, 30MB mem, 98% CPU, 2GHz)
* C.pm: MSVC fix for "(OP*)Nullany" i.e "(OP*){0}" => "{0}". Fixes
test 27 on MSVC. CC test 12 broken on MSVC AS 5.8
* bytecode.pl: Restore C strictness, printf before declaration.
Undefine debop for MSVC also. Fixed [cpan #53228].
* t/test.pl: Cleanup intermediate MSVC .obj. Skip too slow tests 29
on MSVC, only c.t
Remove >nul for test results on Win32.
More run_cc_perl fixes for Win32.
Be more quiet on MSWin32 also (no 2>/dev/null)
Skip stderr redirection to stdout on cmd.exe tests, t/o.t 5,9.
* t/bytecode.pl: ditto.
* t/TESTS: Fixed 28 for MSWin32 MSVC
1.12 2009-12-31 rurban
__DATA__ handles work now. Also more Bytecode fixes.
* t/testplc.sh: Added -S: skip all but -S and Concise
* Bytecode.pm (1.05): Adapt RV 5.11 change from SV to IV, detect ROK at
B::IV::bsave analog to C.pm. Fixes test 16.
Set prototype sv_flags at first, so it cannot assert. Fixes tests 27 and 29.
More work on 5.6 compat.
* bytecode.pl: Fixed 5.6 %alias_to sizes.
* bytecode.h: Fix BSET_cop_label for 5.11. Fixes test 21.
* t/bytecode.t: Added DEBUGGING to the mix. 5.11 is now completely the
same as 5.10. Remaining failures are now 9,10,12 for DEBUGGING (or
non-threaded?).
* C.pm: Added xpvuv (coverage in 5.6.2 test 21).
Added PMOP->pmstash/pmstashpv (no coverage yet).
Checked against -Wall -Wextra:
Added missing pmop->stash for 5.8.
Added missing cop->cop_io for 5.8.
Added braces to union inits (MSVC Todo).
Remove unused dTARG and dSP.
Fixed CC test 29 on 5.6.2, not checked why.
Fixed __DATA__ test 15 on 5.10 and 5.11.
* TESTS: Updated with 5.10.1 CORE results
1.11 2009-12-29 rurban
Fixed 5.11 RVs, added C optimizations to CC
* C.pm: Adapt RV 5.11 change from SV to IV, detect ROK at B::IV::save.
Fixes test 16.
Added -DW for walk, together with -Dp only.
Forced fullname optimization scalars for B::CC.
Export save_sig and fixup_ppaddr for CC.
* CC.pm (1.03): Fixed Perl_gv_SVadd cast warning.
Added most B::C optimizations. This fixes all threaded tests 15 >= 5.10.
Call save_sig, fixup_ppaddr not yet.
* t/test.pl: simplified (and fixed) keep logic
* t/c_argv.t: cleanup
1.10 2009-12-29 rurban
Fixed @ARGV
* C.pm: fixed couple of minor bugs detected with new t/testcore.t:
t/CORE/comp/retainedlines.t => Use of uninitialized value $len, save_hek
Use of uninitialized value $i in printf at C.pm line 49
Undefined value assigned to typeglob at C.pm line 2901
Do not overwrite run-time @ARGV on >= 5.10 (GP and AV). Fixes test c_argv.t
* t/testcore.t: added
* t/c_argv.t: added. test @ARGV and perlcc
* perlcc (2.06): grab_stash only via new -stash, not automatically;
B::C is better.
ldopts -std only when modules are given.
fix for empty options.
add -sharedxs, -sharedlib; -shared is now for shared linkage,
-sharedlib to create a shared lib (was -shared before but never worked),
-sharedxs to also link shared xsubs, no run-time DynaLoader/XSLoader needed.
1.09 2009-12-27 rurban
* C.pm: Cut down the number of AUTOLOAD'ed optrees, search only on
empty gv->cv (31).
Added very unsafe -O4/-fcop to eliminate most cops and file + line numbers.
Added faster -fav-init to -O2, less unneeded av_extend().
Get rid of fbm_compile (11). Only 5.8.9 is broken.
* CC.pm (1.02): partially fix eval "" for >=5.8.9, wrong CORE cc_runtime.h
definition. (test 29 - eval "")
* t/test.pl, t/c*.t: Simplify C/CC testsuite.
* t/testc.sh: fixed -E
Added -o for orig, no -Mblib
* t/c_o3.t, t/c_o4.t: added
1.08 2009-12-23 rurban
Adapt coding guidelines, add C and cperl coda. Reformat sources.
* C.pm: Detect AUTOLOAD and CLONE (test 27 and 31).
5.10 is now the best release for unoptimized B::C. 5.8 is faster optimized,
but misses some autoloaded xsubs.
Simplify PVMG exclusion; all non-magical, not just PAD_OUR, as in Bytecode.
* Bytecode.pm (1.04): Add missing BM logic to GV >= 5.10. Fixes test 11.
Fixed op_pmflags logic for 5.10 threaded only, no += 2. Fixes test 2.
Protect empty match from setting op_reflags. Fixes test 20.
Add perlversion to the -S generated header.
Document more options -TI, -TF, -u<package>
* perlcc: improve help. use Pod::Usage as module, not via cmdline.
fix previous installation quirks.
* assemble: improve pod. make it installable.
* disassemble: make it installable.
* cc_harness: ditto.
* Makefile.PL: added script/* as EXE_FILES to install them.
* t/TESTS, t/testc.sh, t/testplc.sh: test 31 non-xs AUTOLOAD added
* ByteLoader/Makefile: tuning, bytecode.h triggers now byterun.c
* bytecode.h: Specialize pregcomp on 5.10; add PL_regex_pad[0] for
pregcomp if missing. fixes tests 3,4,5,7,31 on 5.10.
1.07 2009-12-19 rurban
* C.pm: fix fbm_compile (INDEX) for >= 5.10. gv_list is already a pointer.
rewrite PVGV (ie PVBM) init for >= 5.10. (test 11)
Added CV->save_magic (untested)
Fixed two wrong my decelarations detected by the new Od debugger.
* README: Clarify DIFFERENCES section for CC only. Esp. the
Context of ".." section. New test 30
* t/TESTS, t/testc.sh, t/testplc.sh: test 30 added for Context of ".."
* t/testc.sh: fix -k logic, keep .pl for -O tests
1.06 2009-12-18 rurban
Going public on the Perl birthday
The first public compiler re-release since it was removed from CORE.
* C.pm: Had to disable -O1 -fcog (pv_copy_on_grow) on 5.10 and higher
until I find out how to fool S_unshare_hek_or_pvn. This fixes all C
-O1 and -O2 tests. Warn about this.
Fix wrong RV init on *5.10* for test 29 also, as with 5.11.
Simplify __DATA__ handling. Warn when -O3 or -fsava-data is missing.
Enable working -fsava-data on 5.8 and 5.10 threaded. Fixes couple of tests 15.
Fix 5.8/5.10 PVMG pv = &PL_sv_undef for empty comppadnames
* t/testc.sh: -k keep added. Delete temporaries on PASS as new default.
-E added. Dump preprocessed source file with cc -E as _E.c (was default before)
* Version bump to official:
B::C 1.06 (1.05 is in 5.8.9)
B::CC 1.01
B::Bytecode 1.03
B::Asmdata 1.03
B::Assembler 0.09
B::Disassembler 1.06
B::Bblock 1.03
ByteLoader 0.07
1.04_33 2009-12-17 rurban
* t/cc*.t: skip some memory eating failing tests.
* MANIFEST: dummy lib/B/Asmdata.pm added to fix cpan smokes on BSD.
Thanks to Slaven Rezic.
* C.pm: fix tests 14,16,23 on >=5.10, missing AV magic for
CV main::a, protect PVMG from SvPAD_OUR.
fix wrong RV init on 5.11 for test 29
* CC.pm: fix ARGS >=5.10 (in require, test 28).
Enable quiet. No -q for quiet, just -v (verbose) or any -D switch turns
on verbose.
* t/testc.sh: Added test 27 (import), test 28 (require), test 29 (use)
[Nick Koston]
1.04_32 2009-12-14 rurban
Only two 5.10 B::C bugs remaining!
* C.pm: added missing FBM logic to GV >= 5.10 (test 11).
Forbid MAGIC on non-magival SVs >= 5.10 (tests 14 + 23).
Special PVMG PL_sv_undef string. (namesv problem on 10,14,23)
1.04_31 2009-12-14 rurban
* C.pm: fix test 21 on 5.10 "Bad free() ignored". Dynamically
allocate CopLABEL.
* Makefile.PL: fix the make depend section > 5.9
* t/testc.sh: fix -D option
1.04_30 2009-12-13 rurban
Better 5.6 support, but Bytecode compiler still disabled.
Not even the Byteloader works for 5.6 assembled code. The Disassembler
does not stop at ret. Fixed several C tests.
* bytecode.pl: renumbering policy: <= 5.6 squeeze, >= 5.8 fixed
to force exact bytecode numbering as with 5.8.
Allow float versions, such as 10.001 for subreleases.
Add perlversion to header.
Add byteorder check logic.
Added logic to include all possible unsupported ops, all holes.
bytecode table checked for 5.6.1, 5.6.2, 5.8.0, 5.8.1-5.11
Note, that I found official and contradicting 5.6.1 releases.
ActiveState 5.6.1.626 had no op_children, the unix release tarball
with 5.6.1 (called Trial 1) still has it.
* ByteLoader.pm (1.06_06): better bytecode numbering.
* Asmdata.pm (1.02_02): strict numbering along 5.8.
Unsupported ops are now also included, with PUT method 0.
* Assembler.pm (0.07_07): Add perlversion to header
* Disassembler.pm (1.05_04): Add perlversion to header.
Print header only for existing.
* BcVersions/Asm*.pm, BcVersions.pod: added for most versions with the
CORE compiler.
* disassemble: Make it work with the older Disassembler.
Make it work with older bytecode >= 5.8 via Asmdata unsupported ops.
* i386.xs, jitcompiler.pl: removed (seperate module)
* bytecode.h, C.xs: Fixed for 5.6, tested with 5.6.1 with MSVC.
* bytecode.h, byterun.c: add static int force = 0. Override it to 1 to
try out executing older bytecodes.
* C.pm: fix test 16 on 5.10 again, non-threaded only (GVAV magic).
first fix in 1.04_24, broken in 1.04_28. On 5.11 I have a broken
RV handling here.
Remove version and Regexp::DESTROY from the automatic methods
(experimental).
First attempts with pv and hek consting failed.
* t/testc.sh, t/testplc.sh: support proper cmdline arguments, add -h for
help
Fixed test 5 by fixing the wrong test.
* perloptree.pod: renamed from perloptreeguts.pod. Added links.
1.04_29 2009-12-07 rurban
Disable the Bytecode compiler for 5.6. Fixed besides the __DATA__ handle
(test 15) all remaining 5.8 C bugs.
* C.pm: assign the PVX of GV also for empty stashes (main::) to avoid
destruction failures at Perl_pad_undef() SvPVX_const(namesv) == '&'.
Fixes tests 14 and 23
* byterun.c: No assert on empty bs_sv, it is &PL_sv_undef.
Map missing -Dv on 5.6 to -Df i.e. -D256
Added better debugging output: setter values.
* STATUS: re-run CORE tests, list cmdlines, update status
* Disassembler.pm: added gen_header_hash() for gen_header output.
add to documentation.
* Bytecode.pm: -S uses gen_header_hash for gen_header output.
1.04_28 2009-12-06 rurban
* C.pm: fix -fcog (Copy on Grow with -O1 and -O2) for >= 5.10 by a new
SVPV.
Statically initialize PVs with a union layout trick. Use char* as first
union member.
Revert xpvcv+svsect changes to pre-initialize those structs against
gvcv recursion.
* t/b.t, t/o.t, t/stash.t: fix tests for 5.6.2
* bytecode.c: byteorder swab only if same length.
1234 => 4321, 12345678 => 87654321.
Skip a 0x prefix for byteorder in the header. Perl 5.6 had this.
* Assembler.pm (0.07_06): 5.6 byteorder fix: until 5.6 the
$Config{byteorder} was falsely dependent on ivsize, we need longsize.
Add 0x to byteorder in header again, as with 5.6.
* t/testc.sh: do not run old leftovers. some 5.6 optimizations
* t/c_o1.t, t/c_o2.t, t/cc_o2.t, t/cc_o2.t: added
* t/test.pl: support -O1 and -O2
* broke ByteLoader
1.04_27 2009-12-04 rurban
* CC.pm (1.02_03): fixed endless loop in optimized sort. The sortcv
cmp needed a return NULL (test 19).
* c.t, cc.t: add failing tests as TODOs
1.04_26 2009-12-02 rurban
* C.pm: remove nv->IVX for 5.10 (Can't locate object method "IVX" via
package "B::NV").
protect mg->OBJ when SCALAR (Can't call method "save" on unblessed
reference).
fix NV isa IV, to fix test 23 (Nick Koston).
fix CopLABEL for 5.11 (test 21)
fix PMOP extflags, esp. for split on >= 5.10 (test 5 + 7)
add fbm_compile() to BM::save >= 5.8 (test 11) (Nick Koston)
* t/c.t, t/testc.sh: added tests 23, 24 (Nick Koston)
* Bytecode.pm (1.02_04), bytecode.h: store and set PMOP reflags
as U32. Fixes tests 2-4,7
Start of backporting Perl 5.6
1.04_25 2009-11-30 rurban
* C.pm: fixed IO for 5.11: LINES => IVX.
specialize GvLINES = -1 and protect against U32 signed-ness.
fixed test 8 main::AUTOLOAD (unoptimized only)
* t/test.pl: enable -Mblib for all Perl versions.
fix 5.6.2 ldopts by explicit checking of (-lperl|CORE\/libperl5).
make test works now for 5.6.2
* t/cc.t: adapt TODOs
* t/c.t: add TODOs 15 for 5.6
* script/perlcc: a.exe for cygwin also
1.04_24 2009-11-29 rurban
* C.pm: fixed more CV basics (test 9), just stash for autoload missing
(test 8).
PVLV adjusted, but no testcase yet.
IO adjusted. Fixes test 4.
re-enabled Perl 5.6, all tests but 8,15 pass with t/testc.sh. make test not.
added option -DO, more OP info than with -Do.
added saving "main::" CVs. Fixes test 16.
Fixed PMOP op_pmreplroot for >= 5.10 (for SUBST). No test
* t/TESTS, t/testc.sh: added 20 for qr// bug in 5.10
* t/c.t: remove TODO status from tests 14-16
1.04_23 2009-11-22 rurban
* yapc_bratislava08.pod: title fix
* SIGNATURE: added
* C.pm: GvGP handling improved. added more sect comments.
added -u option "unused" not to print verbose debugging info
for unused subs.
Fixed CV XPVCVIX initialisation - off-by-one error, wrong CvFLAGS.
Fixes test 6.
* run perltidy -i=2
1.04_22 2009-11-16 rurban
* C.pm, CC.pm: fixed illegal pad in ASSERT_CURPAD_ACTIVE("pad_sv")
with DEBUGGING. Simply forgot to set PL_comppad
1.04_21 2009-11-07 rurban
* yapc_bratislava08.pod: added
* C.pm: fixed CALLREGCOMP >=5.10 by using dynamic newSVpvn as in ByteLoader
1.04_20 2008-06-25 rurban
* t/TESTS: numbered.
* t/TESTS: added sub FETCHSIZE to test 16. required now
* C.pm, CC.pm: change debug globals to hash.
* C.pm: Fix B::RV::save for 5.10
* CC.pm: 5.10 fixed GvSV to GvSVn (PERL_DONT_CREATE_GVSV), fix cc test 3+4 crash
* Artistic, Copying: added. Clarified perl license.
* Bytecode.pm: fixed IO=>PVNV problem in test 15.
* bytecode.pl, bytecode.h: rewrote gp_sv and gp_file setters
with x (special setters). Fixes unshare hek assert in test 15, and force
creation of GVSV.
* Assembler.pm 0.07_05: cygwin text-mount fixes with \r\n
* t/testplc.sh, t/testc.sh: require bash, accept testnumber args
* Asmdata.pm: fix PORTABILITY docs
* t/testc.sh: -Bdynamic: link to shared libperl
* cc_harness: simplify. fixed for -Bdynamic -E
1.04_19 2008-06-08 rurban
* CC.pm: CXt_LOOP check rewrite, use now CxTYPE_no_LOOP()
* CC.pm, t/*.t: useithreads, not usethreads
* bytecode.t, test.pl: TODO handling: do not print "ok 1 #"
* b.t: fixed skip logic, better skip message
* Bblock: fixed to work without B::Concise (5.6.2)
* C.xs: add PM_GETRE for 5.6.2 and earlier
* CC.pm: Fix PP_EVAL hack from 1.04_18
* Bytecode: fixed IO::SUBPROCESS error (test 15)
* cc_harness: check static libperl.a, else try dynamic. Support -E
* ByteLoader: support different byteorders (Disassembler not yet)
* ByteLoader, Assembler, bytecode.pl: add archflags to header, relax
platform strictness.
* t/stash.t fixed for 5.10
* Disassembler: added hashref return for get_header in scalar context.
* Bytecode: fixed compiler crash/assertions for GV without GP (test 11)
1.04_18 2008-06-04 rurban
* Fixed t/assembler.t on 5.6.2
* Added ByteLoader::unimport (from 5.6.2)
* ByteLoader: Fixed bytecode crashes for 5.10+5.11 in 2-5,7 by
removing PMf_COMPILETIME from op_pmflags. Reason: empty
check_substr in INTUIT. (6 still failing)
* ByteLoader: Fixes for 5.11 by using PM_SETRE (test 7 still failing)
* Bytecode: removed op_reflags. pregcomp with op_pmflags alone
is enough to reconstruct it.
* C,CC: enabled static_ext + DynaLoader boot section again. Fixed in
cygperl.
* CC: fixed Bug#55302 PP_EVAL by adding PP_EVAL_thr for threaded perl.
cc_runtime.h was always wrong on this!
1.04_17 2008-04-20 rurban
Added Fedora Core 1 as test system (5.6.2, 5.8.3, 5.11.0@33708 + valgrind)
* ByteLoader: Fixed broken BYTEORDER check (for 32int)
* ByteLoader: Remove allowing older versions until we have
an opcode table for older versions.
* ByteLoader: Fixed SIGSEGV with older gcc (3.3.2) at
bstate->bs_obj_list[0] = NULL
* ByteLoader, C.xs: Newx() support and more for older perl's.
Tested with 5.8.3
* bytecode.pl: Fixed <11 op_reflags (double definition)
* bytecode.pl: Support 8-10 version syntax (op_pmstash, cop_io)
* Makefile.PL: added MKPATH blib/bin
* added ByteLoader/ppport.h
* fixed t/stash.t for 5.11
1.04_16 2008-04-14 rurban
* ByteLoader: Fix PL_tokenbuf for older perls without PL_parser
* B::C: walkoptree_slow for verbose statistics op_count.
* Consistent in -v verbose and not verbose.
* Workaround for invalid pad in save_context.
* Added -Dp for verbose package cache printing.
* -v verbose adds more generated comments: op struct fields
* Fixed cop_label for 5.11 (Change #33656)
* Fixed make subdirs-test error in 5.8.8
* Fixed RT#52920 "invalid cop_free of nullified cop" for B::C
1.04_15 2008-04-11 rurban
Highlights: bytecode and ByteLoader portability preparation and
commented disassembler output. PVGV and new REGEX still broken.
Fixes for no ithreads.
* Fixed verbose Bytecode for old B::Concise (5.8.8). t/testplc.sh
failed before.
* ByteLoader: Work on portability (different arch, version and sizes).
Added bl_header.
Changed macro BYTECODE_HEADER_CHECK to function bytecode_header_check().
Fixes for no threads.
* bytecode.h: pregcomp fix for 5.10 w/o threads
(Thanks to david@cantrell.org.uk for his cpantest for netbsd).
use xpv_cur and not xpv_len for the rx length (len - 1) to strip off
the ending nul byte.
5.10: new minlen check aborts
* ByteLoader security: added strconst maxsize flag to bytecode.pl
for buffer overflow checks.
* Bytecode -DA for devel assertions of absolute or probably wrong opindex
pointers (nyi)
* disassemble, B::Assembler: commented output, similar to -MO=Bytecode,-S
Added backwards option --bare for older assemblers.
* assemble, disassemble: added pod
* B::Disassembler:
Provide default printers print_insn (commented) and print_insn_bare.
Portability: Use the header settings from the .plc and not the current
perl settings.
* B::Assembler 0.07_03, B::Disassembler 1.05_02, ByteLoader 0.06_03:
Added longsize to the bytecode header, just for xcv_depth.
* B::Debug 1.05_03: Fixes for no threads.
* Makefile.PL: Fixed script/perlcc.PL dependency
make: *** No rule to make target `script/perlcc', needed by
`blib/bin/perlcc'.
(Thanks to david@cantrell.org.uk for his cpantest for solaris)
* Added test TODO's, Added perloptreeguts.pod
1.04_14 2008-03-24 rurban
* Fixed some 1.04_13 +x executable permissions in docs. Windows habits.
1.04_13 2008-03-24 rurban
* Fixed B::C pv init in 5.10. Forgot refcnt and flags.
* Added "Other perl to exe compilers" section to perlcompile.pod.
This is even a FAQ.
* Added the NOTES to perlcompile.pod
* Added temp. perloptreeguts.pod. In work, also at the wiki.
* Fixed Bytecode op_pmflags assertions.
* Added orangesect for enough regexp space in RE's sv_any.
Engine not initialized though.
* Hangs in C/CC executables at an op_free(PL_main_root) trying to free
the static op structs because the earlier op_seq -1 hack is gone now,
and opt_static was replaced by opt_latefree. Set now opt_latefree = 1.
This fixed tests 1 and 13. Slowly making progress step by step.
1.04_12 2008-03-16 rurban (first upload)
* Fixed bug #40435 CC: Can't locate object method "_save_common_middle"
via package "B::FAKEOP"
* move tests from [bc]*.t into shared TESTS file.
* fixed remaining ugly Bytecode bug in B::AV::bsave
$_ = $_->ix for @array.
Now we are at the same state as with 5.9.4
* fixed wrong bc cop_warning cast for 5.8
* fixed CC pmreplroot and pmreplstart fields
* make C and CC less verbose, honor -v, -q ignored.
1.04_11 2008-03-09 rurban
* removed Jit and Asm stuff. Maybe as seperate packages somewhen.
But targeting PIR makes more sense.
* added STATUS
* added old Changes from the new git clone repo by Sam Vilain
* fix C cop
* fix C savepvn apparently
* work on HV init - crashes at invalid entry
* use run_cc_test
1.04_10 2008-02-26 rurban
* change bc pvx access
dummy struct bytestate_xpv, so pv has the same fields as with 5.8,
seperated from the sv. The sv is seperated from the pv.
TODO: different strategy for B::C, must link xpv to sv
* add and parse Assembler comments
* add Bytecode comments to -S output unless -q
* change op_reflags from U16 to U32
* guarantee ByteLoader backwards compatibility: Accept lower versions also.
1.04_09 2008-02-25 rurban
* restored 5.8 backwards compatibility, mainly for test comparison.
CPAN installation will abort.
* make t/test*.sh PERL independent
* fixed B::C GV init crashes (SvPOK assertion)
* added B::C RE section
* fixed B::C xpvcv and xpvnv section
* comment Bytecode -S nice ops to re-assemble them without errors
* added new asm opcode op_reflags
* removed Bytecode cop_arybase for 5.10
* 5.8 compatibility: fixed bytecode.pl versioning, Asmdata creation
(exact same for 5.8.8), asmdata.t, b.t tests
* improved make clean
* add ByteLoader -Dt and -Dv debugging
* don't link against jit code against possible sideeffects
* fixed c.t and cc.t tests
1.04_08 2008-02-22 rurban
* fixed t/b.t tests for 5.11 (REGEXP, ref RV => IV)
* fixed Makefile.PL deps to ignore the interim .pl files
* added PMOP reflags
* added madprop to B-1.18_01 and B::C (if provided by B)
* enhanced B::Debug 1.05_02
* fixed B::C PL_cshlen (already initialized since 5.10)
* fixed B::C GvFILE
* fixed various (SV)xpv list casts and inits,
* fixed B::C xpvnv_list for 5.10
* fixed B::C xpvio_list for 5.10
* fixed B::C pmopsect for 5.11
* bootstrap also all $Config{static_ext}, not only dynamic stashes
1.04_07 2008-02-20 rurban
* moved bstate->bs_pv.xpv_pv slot to bs_pv.xiv_u.xivu_p1
* fixed pv_free
* added op_pmflags 2 to match PMOP
* moved B into lib/B
* fixed ByteLoader Makefile deps
* added -DM flag to Bytecode backend
* added perlcompile.pod and Perl_5_Internals.htm
1.04_06 2008-02-19 rurban
* no crashes anymore for 5.10, just op_pmflags & PMf_ONCE assertions.
* blead@32980 crashes in tests 11,16,17,18
* enhanced B::Debug (no version bump)
* fixed general op_list inits, and specials for av, hv
* almost fixed pv within sv handling
* added -O=C,-DS for SV debugging
1.04_05 2008-02-18 rurban
* added t/test.pl and t/test*.sh to MANIFEST.
* fixed ByteLoader reading from the <DATA> filter.
* fixed -H .plc header parsing
* updated Bytecode options in NOTES and pod
* added -O=Bytecode,-v option
* fixed PL_preprocess and SvREPADTMP_off for 5.11
1.04_04 2008-02-08 rurban
* added jitcompiler macros
1.04_03 2008-02-05 rurban
* added Jit and Asm layout, fixed Makefile deps
1.04_02 2008-01-22 rurban (never uploaded)
* removed from CORE, now on CPAN.
* added byteorder to bytecode header.
* added support for 5.10 (NOT WORKING!), 5.9.5 not tested.
up to 5.8.x already in CORE, so disabled.
add back support later, when C/CC is improved or more features
are added.
* added type hekindex
* added c.t and cc.t tests
* extended bytecode.pl format: added version logic
to have strictly consecutive indices
1.04_01 2008-01-15 rurban (never uploaded)
* first package layout for all now non-CORE packages.
===================================================================
B::C was at 1.05 when removed from core with 5.9.4.
Nicholas Clark <nwc10+p5p4@colon.colondot.net> 2007-05-07 15:35:56
bytecode.pl: Exterminate!
ext/B/B/Asmdata.pm: Exterminate!
Nicholas Clark <nwc10+p5p4@colon.colondot.net> 2007-05-07 14:53:05
Given that @optype and @specialsv_name are hard coded tables, it seems
more logical for them to be in B.pm, rather than in the "boilerplate"
for the machine generated B/Asmdata.pm
Marcus Holland-Moritz <mhx-perl@gmx.net> 2004-08-29 13:22:48
Fix typo in B::Assembler.
Marcus Holland-Moritz <mhx-perl@gmx.net> 2004-08-29 13:21:50
Skip ext/B/t/assembler.t when configured without B.
Rafael Garcia-Suarez <rgarciasuarez@gmail.com> 2003-08-09 21:13:27
Remove ByteLoader from the list of modules whose compilability should
be tested : it's loaded (at runtime) by ext/B/t/bytecode.t (as Enache
pointed out.)
chromatic <chromatic@rmci.net> 2002-05-10 07:43:25
added ext/B/t/o.t
Mattia Barbon <mbarbon@dsi.unive.it> 2002-01-12 00:29:48
@14216 B::C, perlcc.PL, B.xs, B.pm, t/TEST, C.xs
Message-ID: <3C3F756C.4581.2E2A938@localhost>
B::C (1.01): added B::C::InitSection, ...
Mattia Barbon <mbarbon@dsi.unive.it> 2002-01-06 12:44:30
@14104 B, B::C, perlcc, t/TEST
Message-ID: <3C38389E.7831.493570@localhost>
Jarkko Hietaniemi <jhi@iki.fi> 2001-12-21 14:42:31
@13830 packing I32 with L is not nice, need l; from Wolfgang Laun.
Jarkko Hietaniemi <jhi@iki.fi> 2001-12-19 16:59:02
@13807 One more embedded (?sx) de-embedded, from Wolfgang Laun.
Jarkko Hietaniemi <jhi@iki.fi> 2001-12-19 15:25:27
@13802 B::Assembler/B::Disassembler patches and test;
from Wolfgang Laun.
TODO: getting perlcc working.
Michael G. Schwern <schwern@pobox.com> 2001-12-14 20:59:13
@13697 BBlock: Some weak docs
Message-ID: <20011215005913.GC28596@blackrider>
Michael G. Schwern <schwern@pobox.com> 2001-12-14 20:14:14
@13695 Adding B::Asmdata docs
Message-ID: <20011215001414.GA25077@blackrider>
Jarkko Hietaniemi <jhi@iki.fi> 2001-11-29 02:22:02
@13346 Change $=, $., $*, $%, and $- to be IVs instead of longs.
Jarkko Hietaniemi <jhi@iki.fi> 2001-11-16 01:38:41
@13034 Add the fruits of Larry Shatzer's version verifying script.
(There are some straddlers, but they will be fixed in the
upcoming releases of the modules.)
(Stamped with v1.00)
Jarkko Hietaniemi <jhi@iki.fi> 2001-09-16 22:36:58
@12040 Stamp B::Assembler with v0.03
Jarkko Hietaniemi <jhi@iki.fi> 2001-07-01 17:20:38
@11061 Make 'compile' target a little less broken.
Still very broken, though: -Wall warnings from
the generated code, boot_Foo prototypes missing,
can't autoload Fcntl::SEEK_CUR et alia, ...
Jarkko Hietaniemi <jhi@iki.fi> 2001-06-27 13:36:40
@10979 "lose the looses", from Abhijit Menon-Sen
Richard Soderberg <p5-authors@crystalflame.net> 2001-06-21 09:01:22
@10780 Step 2: Use the new PM_GETRE/SETRE macros everywhere.
Message-ID: <Pine.LNX.4.21.0106210657400.1693-100000@oregonnet.com>
Gurusamy Sarathy <gsar@cpan.org> 2001-03-19 08:07:09
Subject: Re: sync sync sync: have I missed any patches?
Replace djSP with dSP.
Mark-Jason Dominus <mjd@plover.com> 2001-01-16 14:43:18
Subject: [PATCH @8436] Eliminate op_children
(Replaced by #8448) Traces of op_children (cleanup of #8442)
(Replaced by #8448) More op_children traces (cleanup of #8442).
Subject: [PATCH #3 @8436] Re: Eliminate op_children
Replace #8444 and #8445.
Gurusamy Sarathy <gsar@cpan.org> 2001-01-30 14:20:24
integrate changes#7984,7987,8010 from mainline (gets rid of
dTHR which has been a noop for a while now, except for the
compatibility definition in thread.h)
@8010 remove dTHR;
Gurusamy Sarathy <gsar@cpan.org> 2000-12-18 03:37:02
@7518 UNIVERSAL::can
Gurusamy Sarathy <gsar@cpan.org> 2000-12-18 00:03:38
Fix a couple of compiler-noted nits in #7235.
Fix of sorts for bug id 20000901.092. There seems to be no trace
of a 'pmshort' anywhere in the B, so the offending line was simply
removed.
Gurusamy Sarathy <gsar@cpan.org> 2000-12-17 22:49:13
Subject: [ID 20000928.002] perlcc & ByteCode.pm option mismatch
Did not apply cleanly, manual intervention was needed.
Subject: [ID 20001003.006] B::Debug not -w clean
Test harness update to sync with the new perlcc,
from Simon Cozens.
Gurusamy Sarathy <gsar@cpan.org> 2000-11-27 18:11:21
Bytecompiler patches from Benjamin Stuhl.
More bytecompiler.
Subject: [PATCH blead] B:: missing dependency
Byteloader 0.04
Gurusamy Sarathy <gsar@cpan.org> 2000-11-27 13:53:18
U8->U16 CvFLAGS(cv)
Jarkko Hietaniemi <jhi@iki.fi> 2001-10-25 18:07:58
* make re, Opcode, File::Glob and B threadsafe
* include XS_VERSION in MY_CXT_KEY (tweak for change#12652)
Jarkko Hietaniemi <jhi@iki.fi> 2001-04-05 04:00:33
Integrate changes #9544,9547,9549(perlio),9550,9551 from
maintperl into mainline.
"double" should be "NV"; standard typemap is missing entry
for NV
bailey <bailey@bailey_vms> 2000-02-09 09:09:45
Resync with mainline
Gurusamy Sarathy <gsar@cpan.org> 2000-02-25 02:50:04
@5250 more Compiler tweaks for useithreads
Gurusamy Sarathy <gsar@cpan.org> 2000-02-21 08:02:16
@5178 get Compiler "working" under useithreads
Gurusamy Sarathy <gsar@cpan.org> 2000-02-19 18:33:05
@5150 more B fixups to cope with empty GVs (these can only happen in pads)
Gurusamy Sarathy <gsar@cpan.org> 2000-02-18 04:57:43
@5125 Compiler fixups from Jan Dubois
Gurusamy Sarathy <gsar@cpan.org> 2000-01-27 04:56:48
@4910 various pod nits identified by installhtml (all fixed except
unresolved links)
Gurusamy Sarathy <gsar@cpan.org> 2000-01-26 20:10:26
@4905 s/STOP/CHECK/ blocks
bailey <bailey@bailey_vms> 2000-01-20 00:25:30
@4821 Quick integration of mainline changes to date
Gurusamy Sarathy <gsar@cpan.org> 2000-01-06 11:51:07
@4763 fix various C-backend shenanigans
Gurusamy Sarathy <gsar@cpan.org> 1999-12-08 00:16:21
@4662 typos in change#4546
Gurusamy Sarathy <gsar@cpan.org> 1999-12-03 07:52:50
@4631 support -a switch to append bytecode to an existing file and make
perlcc use it (from Tom Hughes <tom@compton.nu>)
Gurusamy Sarathy <gsar@cpan.org> 1999-12-01 02:00:09
@4602 more complete pseudo-fork() support for Windows
Gurusamy Sarathy <gsar@cpan.org> 1999-11-11 11:32:54
@4546 avoid stash pointers in optree under USE_ITHREADS
Gurusamy Sarathy <gsar@cpan.org> 1999-11-11 07:04:20
@4545 another change towards a shareable optree: avoid pointer to filegv
in COP; revert parts of change#4485 and s/xcv_filegv/xcv_file/
(CvFILE() may yet come in handy somewhere); adjust compiler doodads
to suit
Gurusamy Sarathy <gsar@cpan.org> 1999-11-04 19:25:45
@4516 change#4485 didn't do the right thing for B::Bytecode
Gurusamy Sarathy <gsar@cpan.org> 1999-11-04 18:28:29
@4515 implement STOP blocks and fix compiler to use them (minimally
tested)
Gurusamy Sarathy <gsar@cpan.org> 1999-10-29 08:08:50
@4485 more cleanup: avoid unused knowledge of "file GV" notion in CV and GV
Gurusamy Sarathy <gsar@cpan.org> 1999-10-29 05:00:21
@4484 usurp GVOP slot for new PADOP (one small step to making optree
shareable across interpreters)
Vishal Bhatia <vishal@deja.com> 1999-09-30 01:27:28
@4254 [patch _61] Minor corrections in C.pm
Message-ID: <Pine.LNX.4.10.9909292326280.5599-100000@localhost.localdomain>
Charles Bailey <bailey@newman.upenn.edu> 1999-09-29 04:21:31
@4249 resync with mainline 5.05_61
($handle_VC_problem, @4198: $sv->REFCNT - 1)
Vishal Bhatia <vishal@deja.com> 1999-08-25 02:31:33
@4198 add arenas for managing allocations of remaining xpv*v structures
Message-ID: <Pine.LNX.4.10.9908250031000.11727-100000@localhost.localdomain>
Subject: [PATCH 5.005_60] removing extra ref count (compiler)
Doug MacEachern <dougm@covalent.net> 1999-07-25 17:49:00
@4129 B::clearsym
Vishal Bhatia <vishal@deja.com> 1999-08-11 03:43:28
@4095 applied suggested patch with suitable test to detect MSVC
Message-ID: <GFCJELIOGEENAAAA@my-deja.com>
Subject: compiler on win32
Gurusamy Sarathy <gsar@cpan.org> 1999-09-06 22:16:58
@4092 support bytecode and C backends in perlcc (patch suggested
by Tom Hughes <tom@compton.au>); s/-opt/-noopt/ and make the
C backend the default; describe new switches in pod; introduce
PERLCC_OPTS and s/COMPILE_TIMEOUT/PERLCC_TIMEOUT/;
s/COMPILE_TEST/HARNESS_COMPILE_TEST/; document these %ENV
entries
Nick Ing-Simmons <nik@tiuk.ti.com> 1999-08-08 15:53:57
@3925 Will now correctly re-call
bootstrap "Foo";
if requested. This should allow build on Win32 and other
platforms where you cannot link to loadables directly.
Nick Ing-Simmons <nik@tiuk.ti.com> 1999-08-07 14:23:55
@3934 Vishal Bhatia <vishalb@my-deja.com>
[PATCH 5.005_60] fix for some obscure bugs (compiler)
Message-ID: <HLPEBPHPAKHKAAAA@my-deja.com>
Nick Ing-Simmons <nik@tiuk.ti.com> 1999-08-07 14:19:46
@3933 B::C changes to get simple Tk app. compiling again
Nathan Torkington <gnat@frii.com> 1999-08-05 23:25:47
@3927 Fix for Nathan's fix (#3920) from Guy Decoux.
Subject: Bugfix for my bugfix
Message-ID: <14250.21947.765134.940583@localhost.frii.com>
Nathan Torkington <gnat@frii.com> 1999-08-04 13:24:56
@3920 [5.005_60 PATCH] Make B::Bytecode work
Vishal Bhatia <vishal@deja.com> 1999-07-31 21:08:33
@3869 "use constant" in the compiler
Nathan Torkington <gnat@frii.com> 1999-08-01 11:23:35
@3859 ext/B/B/Disassembler.pm patch
Message-ID: <14244.33431.739419.806927@localhost.frii.com>
Gurusamy Sarathy <gsar@cpan.org> 1999-08-01 20:34:41
@3848 fix defined(@foo) encarpments
Gurusamy Sarathy <gsar@cpan.org> 1999-07-27 05:56:17
@3784 change#3762 wasn't needed
Gurusamy Sarathy <gsar@cpan.org> 1999-07-26 10:59:47
@3762 setstate stop-gap from Vishal Bhatia
Gurusamy Sarathy <gsar@cpan.org> 1999-07-26 10:06:39
@3761 patch for pp_foo -> Perl_pp_foo changes from Vishal Bhatia
Gurusamy Sarathy <gsar@cpan.org> 1999-07-17 22:43:27
@3687 make CC.pm use a distinct CCPP() macro rather than PP()
(suggested by Vishal Bhatia <vishalb@my-deja.com>)
Vishal Bhatia <vishal@deja.com> 1999-06-30 16:02:42
@3644 Compiler and XSUBS
Stephen McCamant <smcc@mit.edu> 1999-06-25 15:38:44
@3637 Eliminate CONDOPs
Vishal Bhatia <vishal@deja.com> 1999-06-20 19:17:17
@3630 Minor bug fix in pp_require
Vishal Bhatia <vishal@deja.com> 1999-06-12 10:23:59
@3622 applied patch after demunging headers with appropriate paths
Message-ID: <JIHEJPFDFKIBDAAA@my-deja.com>
Subject: [Patch 5.005_57] unsigned arithmetic (Compiler)
Vishal Bhatia <vishal@deja.com> 1999-06-05 10:42:17
@3593 applied parts not duplicated by previous patches
Message-ID: <JAMCAJKJEJDPAAAA@my-deja.com>
Subject: Fwd: [PATCH 5.005_57] consolidated compiler changes
Vishal Bhatia <vishal@deja.com> 1999-06-03 02:57:48
@3584 [PATCH 5.005_57] pp_sort sorted out
Gurusamy Sarathy <gsar@cpan.org> 1999-06-10 10:38:00
@3526 fix small nits (ARGS => aTHX)
Gurusamy Sarathy <gsar@cpan.org> 1999-06-02 06:47:10
@3518 remove _() non-ansism
Tom Hughes <tom@compton.nu> 1999-05-27 01:59:49
@3499 ByteLoader mark 2
To: perl5-porters@perl.org
Message-ID: <bf337a0849.tom@compton.compton.nu>
plus resolve tiny conflict with #3479
plus regen_headers.
Gurusamy Sarathy <gsar@cpan.org> 1999-05-11 11:34:13
@3367 various fixes for clean build and test on win32; configpm broken,
needed to open myconfig.SH rather than myconfig; sundry adjustments
to bytecode stuff; tweaks to DYNAMIC_ENV_FETCH code to make it
work under win32; getenv_sv() changed to getenv_len() since SVs
aren't visible in the lower echelons; remove bogus exports from
config.sym; PERL_OBJECT-ness for C++ exception support; null out
IoDIRP in filter_del() or sv_free() will attempt to close it
Gurusamy Sarathy <gsar@cpan.org> 1999-05-10 09:49:26
@3359 more Compiler patches from Vishal Bhatia <vishalb@my-dejanews.com>
Date: Tue, 27 Apr 1999 23:47:24 PDT
Message-ID: <19990428064724.95244.qmail@hotmail.com>
Subject: [PATCH 5.005_56] Saving Tied hashes ( C.pm)
--
Date: Thu, 29 Apr 1999 18:21:06 -0700
Message-ID: <GEFPBFDJADFJBAAA@my-dejanews.com>
Subject: [PATCH 5.005_56] double constants ( C.pm)
--
Date: Mon, 03 May 1999 20:21:31 PDT
Message-ID: <19990504032131.81113.qmail@hotmail.com>
Subject: [PATCH 5.005_56] Overloading implementation ( Compiler)
--
Date: Thu, 06 May 1999 17:57:09 -0700
Message-ID: <FCJELBLAJBOBAAAA@my-dejanews.com>
Subject: Stash.pm
Tom Hughes <tom@compton.nu> 1999-04-24 20:11:59
@3356 applied suggested patch, with win32 and PERL_OBJECT additions
Message-ID: <609bdff748.tom@compton.compton.nu>
Subject: ByteLoader patch
Gurusamy Sarathy <gsar@cpan.org> 1999-05-06 10:01:23
@3314 compiler fixes from Vishal Bhatia <vishalb@hotmail.com>
Date: Tue, 30 Mar 1999 23:40:34 PST
Message-ID: <19990331074034.6117.qmail@hotmail.com>
Subject: [PATCH 5.005_56] pp_entersub and pp_leavewrite(CC.pm)
--
Date: Wed, 07 Apr 1999 00:28:23 -0800
Message-ID: <FGBNLNPOEELFAAAA@my-dejanews.com>
Subject: [PATCH 5.005_56] function prototypes(B.pm)
--
Date: Thu, 22 Apr 1999 23:40:52 -0700
Message-ID: <OEAOMKBMLDADCAAA@my-dejanews.com>
Subject: [PATCH 5.005_56 ] discarding worthless padsvs
--
Date: Tue, 27 Apr 1999 01:14:49 PDT
Message-ID: <19990427081449.28615.qmail@hotmail.com>
Subject: [PATCH 5.005_56] pp_ncmp implementation ( CC.pm)
Vishal Bhatia <vishal@deja.com> 1999-03-23 17:21:43
@3168 applied needful parts of suggested patch
Message-ID: <OOOKJGFMLFLHBAAA@my-dejanews.com>
Subject: [PATCH 5.005_56] pp_formline correction
Gurusamy Sarathy <gsar@cpan.org> 1999-03-25 07:21:05
@3165 fix refcnt on PL_main_cv (variant of suggestion by Vishal Bhatia
<vishalb@hotmail.com>)
Vishal Bhatia <vishal@deja.com> 1999-03-11 08:40:27
@3142 applied suggested patch, modulo bogus hunk
Message-ID: <19990311154027.25891.qmail@hotmail.com>
Subject: [PATCH 5.005_56] Minor fixes in perlcc
Vishal Bhatia <vishal@deja.com> 1999-03-03 00:27:25
@3066 updates to compiler modules
Message-ID: <19990303072725.779.qmail@hotmail.com>
Subject: PATCH 5.005_56 + Test procedure
Vishal Bhatia <vishal@deja.com> 1999-02-08 02:14:41
@2940 enable dynaloading in C.pm-compiled programs (non-conflicting part
of suggested patch)
Message-ID: <DHIDFFOPMEFDBAAA@my-dejanews.com>
Subject: [PATCH 5.005_54] some pending C.pm stuff
Gurusamy Sarathy <gsar@cpan.org> 1999-02-12 12:09:27
@2894 add missing hunk in change#2657 from Vishal Bhatia <vishal@deja.com>
Vishal Bhatia <vishal@deja.com> 1999-01-18 02:32:31
@2657 pp_mapstart & pp_grepstart return val (CC.pm)
To: perl5-porters@perl.org
Message-ID: <19990118093231.18443.qmail@hotmail.com>
Vishal Bhatia <vishal@deja.com> 1999-01-17 17:41:10
@2639 Duplicate saved ops (CC.pm)
Message-Id: <19990118004111.29667.qmail@hotmail.com>
Gurusamy Sarathy <gsar@cpan.org> 1999-01-17 10:12:42
@2621 fix change#2602 to not used hard coded constants
Vishal Bhatia <vishal@deja.com> 1999-01-11 09:02:41
@2602 B::MAGIC::PTR doesnot check for valid length.
Lines: 134
Message-ID: <MLIST_19990111052126.27966.qmail@hotmail.com>
Vishal Bhatia <vishal@deja.com> 1999-01-08 13:43:36
@2600 B::CC::pp_rv2cv problem
To: perl5-porters@perl.org
Message-ID: <MLIST_19990108101557.4481.qmail@hotmail.com>
Vishal Bhatia <vishal@deja.com> 1998-12-16 04:17:03
@2525 More minor Fixes in CC.pm/C.pm
To: perl5-porters@perl.org
Message-ID: <MLIST_199812160055.QAA06272@f10.hotmail.com>
Vishal Bhatia <vishal@deja.com> 1998-12-10 09:30:02
@2524 Re:perlcc -e 'my $x = shift; print +($x ?...' failure
To: perl5-porters@perl.org
Cc: rmb1@cise.npl.co.uk, nick@ni-s.u-net.com
Message-ID: <MLIST_19981210061651.29891.qmail@hotmail.com>
(Nick's part was applied earlier, in change #2460)
Vishal Bhatia <vishal@deja.com> 1998-12-09 07:50:30
@2523 return value of perlcc/B::CC generated functions.
To: nick@ni-s.u-net.com
Cc: perl5-porters@perl.org
Message-ID: <MLIST_19981209043146.16829.qmail@hotmail.com>
Vishal Bhatia <vishal@deja.com> 1998-12-20 20:03:25
@2492 pp_next/pp_last/pp_redo problems
To: perl5-porters@perl.org
Message-ID: <19981221030326.27660.qmail@hotmail.com>
(slightly reformatted)
Nick Ing-Simmons <nick@ing-simmons.net> 1999-01-02 14:06:30
@2551 Export constant subs from B.xs for op.h, cop.h and a few others.
Use them in various B::* rather than have local defs.
Nick Ing-Simmons <nick@ing-simmons.net> 1999-01-02 10:04:02
@2550 Integrate ext/B changes from //depot/cfgperl
'copy in' ext/B/B.xs
(@2460..) 'merge in' ext/B/B/C.pm ext/B/B/CC.pm
ext/B/B/Stackobj.pm (@2524..)
Nick Ing-Simmons <nick@ing-simmons.net> 1999-01-02 10:04:02
@2491 Integrate ext/B changes from //depot/cfgperl
Gurusamy Sarathy <gsar@cpan.org> 1998-09-23 09:52:46
B::Asmdata define PUT_svindex(), PUT_opindex()
Nick Ing-Simmons <nick@ing-simmons.net> 1998-12-20 14:21:29
@2491 Save _all_ GV's which have SV, AV or HV set.
Vishal Bhatia <vishal@deja.com> 1998-12-09 23:16:50
@2461 1. Fixes the bug reported by Robin Barker <rmb1@cise.npl.co.uk>
2. Fixes the bug regarding return value of c-functions generated out
of perl subs. ( Just includes the patch I sent earlier)
3. Incorporates the other changes that need to be done to get CC.pm
use ISA search for packages and methods on the same lines as C.pm
Vishal would appreciate comments about B::Stackobj changes from
someone knowing that module well.
Nick Ing-Simmons <nik@tiuk.ti.com> 1998-12-05 17:14:42
@2453 Avoid hard-coding op numbers
Update CC.pm to save %INC, and to co-exist with new C.pm
Nick Ing-Simmons <nik@tiuk.ti.com> 1998-12-05 11:44:28
@2452 B.xs had its own code to calculate hash() which differed from
PERL_HASH in hv.h - so all saved HV's were mangled - including %INC
which meant that run-time require was re-done.
Removed some debug from C.pm
Nick Ing-Simmons <nik@tiuk.ti.com> 1998-12-04 22:58:49
@2451 Snapshot of re-worked B::C which compiles Tk apps at least as
well as _54, but with pre-scan for classes and save the ISA scheme.
Vishal Bhatia <vishal@deja.com> 1998-12-04 18:58:44
@2450 Vishal Bhatia's patch as a basis.
Nick Ing-Simmons <nik@tiuk.ti.com> 1998-11-28 23:46:57
@2378 More C.pm tweaks
Save globs even if we have saved cv itself before - may be imported.
While we don't save "bootstrap" CV we need to provide a stub,
so that if we require it later we don't fall through and attempt
to DynaLoad module again.
Attempt to save %INC so that "require" does not reload things
we have compiled-in (does not work right yet - seems to be due
to PL_incgv being created in perl_parse() current scheme setting
GvHV() is "better" than saving the glob, but still does not
work as I expect).
Nick Ing-Simmons <nik@tiuk.ti.com> 1998-11-27 22:10:27
@2332 Handle INIT list in C.pm
1. Provide init_av() from B.xs
2. Export it in B.pm
3. Use it in C.pm
Also disable some pruning in savecv() which seems to undo
my previous patch.
Experimental feature - save pathnames of .so files in easily
grep-able form for use in wrapper to feed to linker.
Gurusamy Sarathy <gsar@cpan.org> 1998-11-27 15:41:38
@2324 B::C tweaks to allow Tk compiles from Nick Ing-Simmons
Vishal Bhatia <vishal@deja.com> 1998-10-29 00:45:32
@2146 hand-apply whitespace-mutiliated patch
Message-ID: <19981029074534.2334.qmail@hotmail.com>
Subject: [PATCH 5.005_52]Compiling modules,more bugfixes for B
Vishal Bhatia <vishal@deja.com> 1998-10-22 00:59:03
@2072 implement C<goto &func> and other fixes (via private mail)
Message-Id: <19981022055904.20083.qmail@hotmail.com>
Subject: [PATCH 5.005_52] More fixes for B
Vishal Bhatia <vishal@deja.com> 1998-10-11 20:41:38
@1993 fix bug in B::CC::pp_sassign()
Message-ID: <19981012014139.19614.qmail@hotmail.com>
Subject: B::CC problems with pp_sassign routine
Jarkko Hietaniemi <jhi@iki.fi> 1998-10-17 15:17:19
@2005 Integrate mainperl.
Gurusamy Sarathy <gsar@cpan.org> 1998-09-23 11:52:46
@1842 define PUT_svindex(), PUT_opindex()
Gurusamy Sarathy <gsar@cpan.org> 1998-08-02 07:09:35
@1708 fixes for pod noises
Gurusamy Sarathy <gsar@cpan.org> 1998-07-22 02:08:00
@1618 fix up B modules for PL_* changes
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-07-21 18:13:16
@1617 Compiler docs for 5.005
Message-Id: <199807211713.SAA20735@sable.ox.ac.uk>
Gurusamy Sarathy <gsar@cpan.org> 1998-07-22 01:29:09
@1616 s/PL_sv/PL_bytecode_sv/ etc., so we have unique, case-insensitive
names
Gurusamy Sarathy <gsar@cpan.org> 1998-07-21 07:12:00
@1611 fix bytecode.pl with moved var names
Gurusamy Sarathy <gsar@cpan.org> 1998-07-21 05:31:13
@1608 part 2 of PERL_OBJECT fixes (globals in bytecode.h moved to intrpvar.h)
Gurusamy Sarathy <gsar@cpan.org> 1998-07-21 05:29:10
@1607 part 1 of PERL_OBJECT fixes for new var names
Gurusamy Sarathy <gsar@cpan.org> 1998-07-20 09:38:39
@1578 complete s/foo/PL_foo/ changes (all escaped cases identified with
brute force search script). Result builds and passes all tests on
Solaris. win32 and PERL_OBJECT are still untested.
Nick Ing-Simmons <nick@ing-simmons.net> 1998-07-18 13:53:03
PL_ prefix to all perlvars, part1
Builds and passes all tests at one limit i.e. -DPERL_GLOBAL_STRUCT
Ilya Zakharevich <ilya@math.ohio-state.edu> 1998-07-15 06:10:36
@1548 Minor improvements to perlcc
Message-Id: <199807151010.GAA11270@monk.mps.ohio-state.edu>
Gurusamy Sarathy <gsar@cpan.org> 1998-07-15 10:01:41
@1516 add stub docs for ext/B, other minor tweaks
Gurusamy Sarathy <gsar@cpan.org> 1998-07-09 05:37:48
get it building again on win32
Gurusamy Sarathy <gsar@cpan.org> 1998-07-04 05:02:01
@1292 fix perlcc to not rm output file, and other -w(arts)
Stephen McCamant <alias@mcs.com> 1998-06-22 21:19:43
@1200 Inheritance of B:: classes
Message-Id: <m0yoIgR-000EP2C@alias-2.pr.mcs.net>
Gurusamy Sarathy <gsar@cpan.org> 1998-06-19 17:22:23
update repository copy of Asmdata.pm after `perl bytecode.pl`
Gurusamy Sarathy <gsar@cpan.org> 1998-06-11 02:59:23
@1113 fix outdated bytecode.pl
Nick Ing-Simmons <nick@ing-simmons.net> 1998-05-14 18:09:01
@972 Changes to allow compiler with gcc-2.8.1 in C++ mode,
Remove K&R style functions, avoid struct/typedef clash.
Ed Peschko <epeschko@den-mdev1> 1998-04-29 21:02:36
@963 [ PATCH 5.004_64 ] Integrated regression tests for compiler
added perlcc.PL
Ilya Zakharevich <ilya@math.ohio-state.edu> 1998-04-08 03:21:03
@942 Subject: [PATCH 5.004_64] Cryptic error from B::CC
Date: Sat, 11 Apr 1998 19:52:25 -0400 (EDT)
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-04-14 16:22:51
@888 [compiler] CC did "<<" instead of ">>" for right-shift on ints.
Charles Bailey <bailey@newman.upenn.edu> 1998-03-02 01:48:27
Miscellaneous minor fixes
Spider Boardman <spider@orb.nashua.nh.us> 1998-03-01 02:02:47
Almost OK: 5.004_61 (threads, perlio)
Andrew Cohen <cohen@andy.bu.edu> 1998-02-27 20:03:29
Change getc/fread to PerlIO_getc/fread in bytecode.h:
Subject: [PATCH 5.004_61] bunch of small patches
Nick Ing-Simmons <nick@ing-simmons.net> 1998-02-28 11:31:15
Missed FREAD in bytecode.h
Cannot export svref_mutex in non-threaded perl
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-02-27 18:35:27
Change FREAD/FGETC to BGET_FREAD/BGET_FGETC to avoid clash with
preprocessor symbol on Digital UNIX.
Gurusamy Sarathy <gsar@cpan.org> 1998-02-26 03:56:19
@589 [win32] various cleanups so that B can be built as "just another extension"
- export symbols needed for building B
- bset_obj_store() is needed by byterun(), so define it there instead
of at B.xs, and export it
- freadpv() is only used in B.xs, so move it there
- byte*.h are now included by perl.h
- regenerate embed*.h
Gurusamy Sarathy <gsar@cpan.org> 1998-02-22 02:40:56
@568 [win32] get compiler building under win32 (needed Makefile.PL
hacks that could be applicable to other platforms)
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-02-25 17:44:34
@585 More compiler tweaks.
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-02-20 18:23:47
@565 Remove compiler files from their old lib/B locations. The compiler
now builds by default (without the byteperl executable so far) and
seems to work at least minimally.
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-02-20 18:05:33
@564 Move lib/B/... and lib/[BO].pm over to where they should be,
under ext/B.
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-02-20 17:54:58
@563 Start getting compiler to work when built with the core.
[Still won't work as of this change.]
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-02-20 16:42:13
* Merge perlext/Compiler/... into mainline. Some files move to
ext/B/..., some to lib/B/..., O.pm and B.pm go in lib and some
move to the base perl directory (e.g. headers). Will need some
cleaning up before it builds properly, I would guess.
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-02-20 16:39:38
* [compiler] Win32 changes from Sarathy, tweaked slightly by me.
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1998-02-17 17:50:50
* Assorted changes to the compiler
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1997-12-17 10:59:40
* Fix typo in compiler B/C.pm.
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1997-12-10 18:33:53
Alpha5
* Start overhauling compiler. It was working at least minimally
right up until the final tweak of B.xs to add threadsv_names
at which point building it provokes a seg fault in perl while
doing the xsubpp :-(.
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1997-09-03 12:31:48
* Make compiler build/work with devel 5.005
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1997-07-05 11:58:05
* Change %lx to %x in B::CV::save to prevent some CV
fields becoming 0 in the init section. Add missing
write_back in B::Stackobj::Padsv::load_double to fix
test 22 of op/my.t.
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1997-07-05 11:58:05
* B::CC::pp_padsv must cope with vivify_ref (5.004)
as well as provide_ref (5.003)
Malcolm Beattie <mbeattie@sable.ox.ac.uk> 1997-05-03 14:47:06
Alpha 4
* initial check in of compiler
from version Alpha a3 to a4