Revision history for Test-BrewBuild
2.18 2017-06-25
- logging output modifications (changes and relocations)
- in Git::listen(), we now make comparisons of the commit checksums, and
if they are equal, we no longer perform a pull request
- added Test::BrewBuild::Plugin::Author as a required dependency
- added a section to Tutorial that covers using Plugins
- for RPi LCD mode, changed commit csum to 7 chars as opposed to 8,
which is typical/common across the board
- major changes on how we figure out which module is the distribution
name. Instead of taking the first module name found, we check each one
with MetaCPAN::Client, and use the first one found that is a legit
distribution (work on #208)
- fix uninit warning in bbtester. We were trying to use $ARGV[0] even
in cases where it wasn't defined
- fix issue where in Dispatch, we were loading the wrong directive
value into $self->{auto_sleep} from the config file
- fix issue where we getting segfaults when bbdispatch was in auto mode
with --rpi enabled. We were creating a new LCD on each run, and were
eventually hitting wiringPi's MAX_LCD (8) limit. Moved the LCD init
to run once on object instantiation (fixes #209; fixes
stevieb9/rpi-lcd#6)
- in bbdispatch, the --repo|-r flag will prepend https://github.com if
you omit it. To do so, you need to specify the repo in
"user/repo-name" format
- hard coded MetaCPAN::Client version in Makefile to 2.017000 due to a
bug I reported and was fixed promptly by Mickey. See
https://github.com/metacpan/metacpan-client#89
- updated auto log entry at end of dispatch run to include seconds
we'll be sleeping (closes #212)
- fix issue where the repo link was being pre-pended twice with the URL
in log entries due to an earlier change (closes #211)
- opened PR https://github.com/metacpan/metacpan-client/pull/90 for
MetaCPAN::Client
2.17 2017-04-27
- troubleshooting issue where this current commit is causing the LCD
to shift off a couple chars per run. This needs to be looked at
- workaround for above off-by-one LCD issue (stevieb9/rpi-lcd#5). The
global LCD object is now created on each return of a Tester. This
isn't ideal but we'll fix that next
- Git should log out checksums (closes #198)
- each module logs out a characteristic header when an object is
initialized
- cleaned up the log entry for the plugin args in BrewBuild
- added "waiting for tester" log entry line (closes #200)
- re-ordered log entries at the point the tester goes into listen state
(closes #197)
- wrap code lines > 80 cols
- reduced the number of calls to Git::name() in Tester::listen().
Instead of making numerous repeated calls, we now just stash a single
call into a variable (closes #196)
- Tester now logs out its arguments when in debug 7 mode (closes #195)
- POD cleanup (closes #170)
- added RPi integration examples to the FAQ
- slight changes to test_log tests to conform to the changes made within
this version
- Git::git() now keeps a copy of the actual command and returns it
after the first full run. This removes unneeded duplicate log
entries(closes #203)
- several log entry modifications/cleanup
- in Dispatch::auto(), only attempt to fetch the commit csums if there
were results returned (work on #205)
- in Tester::listen(), moved the logging of the csums up one level.
Instead of only logging if we're comparing csums, we log them in all
auto runs regardless if the check is done or not (work on #205)
- added several more command line args to the config file (work on
#202)
- fix issue in Git::revision(). If Dispatch wasn't run from the repo
directory directly, 'git ls-remote' wouldn't ever get any csums back.
Now, revision() requires a repo param if remote is sent in (closes
#205)
2.16 2017-04-21
- remove the named loop in Dispatch::_fork() (closes #188)
- changed all calls to die() to croak() (closes #189)
- in doc and help screen, specify 5.8.9 specifically as the lowest
version of perl we'll use if --legacy isn't sent in (closes #190)
- --auto mode for the dispatcher now defaults to 0 (closes #192)
- fix issue in Dispatch where if we have to set --auto to 0, we were
using the wrong variable rendering early exit
- finished the auto-test (--csum) code in tester/Tester to skip the
checking of repository checksums before deciding whether a test run
should proceed
- now that we've decided not to use plugins for --auto, removed an if()
clause that was in place for that purpose
- in Tester, --csum now is properly passed along when in background
(start) mode
- --auto, --csum and --rpi mode all work correctly
2.15 2017-04-12
- fixed issue in Git::revision() where we weren't actually checking the
remote commit SHA1 sum correctly
- all log objects reflect the module short name, not the full module
path (closes #176)
- moved all auto repo checking and comparisons to Tester
- repo param must be sent in to bbdispatch when running in auto mode
(fixes #176)
- added -a|--auto flag to tester. This must be set when running dispatch
in auto mode (closes #177)
- clean up erroneous and incorrect error in logs about brewbuild args in
auto mode if nothing is returned from the tester
- fix issue with auto, if 0 is specified it wasn't acting like unlimited
but stopping after first run
- fix issue where --auto param wasn't being used when Tester was running
in the background (ie. bbtester start)
- remove Contributing and support sections in POD, updated copyright
year (closes #175)
- added --rpi flag to bbdispatch, used with --auto to enable LCD output
when running RPi::WiringPi unit tests on a Raspberry Pi (this is a
specific purpose feature only for my own testing). In order to work,
bbtester needs to run in auto mode as does dispatch, and the
BB_RPI_LCD env var must be set to a comma delimited list of six pin
numbers that connect the LCD to the GPIO
- added _lcd() in Dispatch for the above addition
- backed out the check in Git::clone() using head(), as it wasn't
working correctly on the RPi
- built in RPi::WiringPi auto tester implemented. Runs tests in auto
mode, and displays results to LCD
- fix bug in Git::revision(). The git command wasn't quoted, so on
Windows, it broke at "C:\Program"
- links within the config file POD now direct to the binaries, not the
libraries (closes #174)
2.14 2017-04-07
- added revision() to Git, returns the SHA1 sum for the most recent
commit, for both local and remotes
- added _separate_url() to Git. This is currently unused as we're not
yet using Github's API to fetch commit info
- added --auto|-a to bbdispatch, spins off an auto run
- added auto() to Dispatch. Checks git commit revision numbers, and if
different, we'll do a pull and run the test suite
- added auto_sleep option to the config file for [Dispatch]. This is how
long in seconds we sleep in between runs and run checks for --auto
- added status() in Git, checks whether the current repo is ahead or
behind in commits compared to remote
- --auto flag to bbdispatch takes an integer, the number of runs to
dispatch out. Send in 0 to run forever
- fully added initial cut of CI automation (via bbdispatch as opposed to
berrybrew) (closes #172)
2.13 2017-03-28
- *NOTE* this release has been tested on win7 with berrybrew v1.10, but
on win10, there may be issues. This is due to an internal library in
that project that is questionable on win10 (Ionic.Zip). This is being
investigated, and an update is planned. I wouldn't recommend trusting
this distribution on Windows 10 at this time for that reason. The
problem stems from the fact that the portable Strawberry Perl that
is being installed is not always extracted fully, despite the fact the
download is ok, and the checksum checks out. Extraction via Explorer
performs correctly. Again, don't trust on win10
- several various POD corrections/updates
- fix issue where info_cache() wasn't being updated after instance
removal (fixes #167)
- added build prereq of Archive::Zip to prevent certain tests from
failing when it isn't installed
- fix bug in BCMD::using() in the Windows regex check. Was missing a \s+
- fix issue in BCMD::info_cache(). A regex was incorrect. It worked on
Unix just due to sheer luck
- added additional argument checking in the command string to Tester
(closes #144)
2.12 2017-03-27
- extended default install timeout to 600 seconds
- fix erroneous uninitialized warning in Dispatch if a repo isn't sent
in, and one isn't found with Git in the cwd. (fixes #165)
- clarify what course of action to take if no testers can be found
when bbdispatch is run (fixes #164)
- added proper links to berrybrew and perlbrew in pod description
(closes #145)
- added LWP::Simple for #142 (check if clone URL exists)
- added check in Git::clone() to ensure that the remote repo exists
before we attempt a clone, and croak() if not (fixes #142)
- added logging to all methods in Git, updated POD, and all locations
that initiate a new Git object send in their own debug level
(closes #166)
- we now cache the results of the '*brew available' calls. Perlbrew
fetches this info from the Internet (and berrybrew will soon too), so
this feature will speed things up when multiple calls for info happen
within a single test run. Can be disabled with the -X param to
brewbuild (closes #163)
- finalized Tester's ability to write logs to a file, and added POD
(closes #149)
2.11 2017-03-26
- we now filter out cperls until we can do further testing. This is
handled in BCMD::available() with a `[^c]` entry in the regex when we
pull out the perls listed in $info (fixes #162)
- _validate_opts() now only compares against argument directives, not
their values (fixes #161)
- in perlbrew v0.77, they introduced a confirmation requirement for
'uninstall', which was breaking things for us. Went through the code,
and found that sending in '--yes' as an argument to perlbrew in the
uninstall command will bypass this confirmation, allowing things to
work correctly here for the perls_remove() functionality
- several test updates due to above changes
- added additional logging to various areas
- bumped prereq of Mock::Sub to 1.07
2.10 2016-11-19
- fix shebang line in the scripts to reference `/usr/bin/env perl`
instead of `/usr/bin/perl` (Thanks 'nigelhorne' (github) for the
report) (closes #157)
- change perms on the bin/ files to 755 from 644 (thanks again to
'nigelhorne' from github for reporting) (closes #156)
- moved BrewBuild object creation in brewbuild down below the checks to
see if *brew is installed. No need for the object if they aren't
(closes #160)
- brewbuild was deleting $args{args} instead of $opts{args} when a
plugin is in use without --revdep (fixes #159)
- in BB::_exec(), we were returning before unloading the dzil shim. This
technically shouldn't have caused any issues, but in some cases it may
have (closes #158)
2.09 2016-11-06
- added "logfile" flag for bbtester. If not in --fg mode with --stdout
set and --logfile is set, we'll write to a file in
"workdir/bbtester.log". This is to help in situations where the
tester is hanging in 'start' mode, but not in --fg mode. Without this
flag, no logging info is displayed or returned (closes #149)
- fixed issue where setup info was not being displayed with -s
(closes #154)
- fixed issue where sending -1 to --new was causing brewbuild help to
be displayed (closes #151)
- required Perl version now 5.10 from 5.8, due to that limit in
MetaCPAN::Client
- removed Perl 5.8 tests from Travis
- added help entry for the -S|--save option
- fixed t/test_api/95, as Mock::Sub now has more revdeps (closes #155)
- slight wording change in Tutorial (closes #147)
- brewbuild will now let the user know that no perls are installed
instead of just outputting nothing (closes #150)
2.08 2016-06-23
- changed from using HOMEPATH env var to USERPROFILE on Windows, as the
former didn't contain the drive letter (fixes #96)
- legacy perls were not being excluded when legacy was disabled
(fixes #137)
- changed all links to dnmfarrell berrybrew repo to stevieb9 (closes
#136)
- removed notices about MSWin PATH issues, as they were completely
rectified in berrybrew after it's "off" feature was added
- several POD corrections
- confirmed that all tests still pass after berrybrew version
sb-20160602 changes (ensured PATH is still ok)
- fixed a couple of tests that had the perl ver hard-coded, and another
two that were trying to use a regex in is()
2.06 2016-06-12
- no functional changes
- BBDEV_TESTING tests now require a version (eg: 5.24.0) set in
$ENV{PERLVER}, instead of having 5.22.1 hardcoded for a few test
files, and other tests don't use the ver anymore (closes #132)
- tests now reside in sub-directories, cleaning the t/ directory up
significantly (closes #133)
- fixed config file doc link in BrewBuild's POD (closes #131)
- new berrybrew build, in prep for using --on on Windows
(https://github.com/stevieb9/berrybrew ver: sb-20160601)
2.05 2016-05-06
- added [brewbuild] section to config file (closes #128)
- cleaned up test logic due to failing MSWin tests
- switched from CPAN::ReverseDependencies to MetaCPAN::Client, as the
former was using MetaCPAN::API which is depricated, and failing to
install on MSWin systems. (I sent a PR to the dist which fixes it). We
now have our own revdep discovery process
- fixed logic issue where -1 wasn't being used in instance_install(),
breaking the ability to install all perl instances (fixes #130)
2.04 2016-04-27
- Results from MSWin now properly show the _bit suffix for the perl used
in test results (closes #124)
- instance_install() when installing randomly doesn't attempt duplicate
versions any longer (fixes #123)
- cleaned up formatting of revdep returned output (closes #126)
- on Unix, we now include Release Candidate perl instances when using
random install, eg perl-5.24.0-RC3 (closes #127)
- various log entry cleanup (closes #112)
2.03 2016-04-26
- brewbuild can now dispatch the most basic of testing to remotes via
the -D and -t flags. For more advanced remote testing, use bbdispatch
(closes #122)
- changed to searching %PATH% for binaries as opposed to using 'where'
on MSWin, as on pre 2k3, 'where' didn't exist (closes #117)
- moved setting the brewbuild command default from bbdispatch to
Dispatch (closes #118)
- On MSWin, we now accept the _32/_64 suffixes in --install, otherwise
we'll default to 64-bit. This suffix will be stripped on Unix systems
if dispatching to both Unix and Windows testers (closes #114)
- removed prerequisite of needing to be an Administrator user on MSWin
from the Tutorial (closes #119)
- fixed a series of out-of-order debug logging message entries
- added a timeout param to instance_install(), default 300 secs. If an
install takes longer than this, it'll roll out the changes and log an
entry of the problem (closes #110)
- enhanced the argument handling for brewbuild to be much more
resilient and specific to help prevent executing rogue commands on
the remote testers (closes #66)
- POD updates
2.02 2016-04-22
- fixed issue in T::BB::Tester where we weren't checking for the 'new'
param, hence we weren't doing install if 'new' was set
- added brief SYNOPSIS in Dispatch
- fixed issue in Tester where the log scalar wasn't being reset after a
run (we also now shutdown() the socket after a finished run)
- correct issues preventing the brewbuild -i and -n flags from working
correctly on remote testers
- dzil dists now get a authordeps run by default, logging organized
(closes #104)
2.01 2016-04-22
- no functional changes
- POD updates/fixes
2.00 2016-04-21
- added T::BB::workdir() as a common source for workdir retrieval
- major POD updates
1.05_02 2016-04-20
- we now set up a log file within Tester for BrewBuild if Tester is
logging, but not to STDOUT
- added a POD document for the config file
- created a Tutorial for the distributed testing environment setup
1.05_01 2016-04-19
- we can now distribute test builds to remote Test::BrewBuild test
servers
- bbdispatch and bbtester scripts added, which handle the test
distribution network/systems
- Dispatch.pm and Tester.pm modules, containing the APIs for the
distribution setup
- we no longer now print anything out by default; everything is
returned, and its up to end-user to handle
- major debug logging improvements
- significant code cleanup and tests implemented
- added/improved upon the flags for the ``brewbuild'' script
1.05 2016-04-04
- we're now Dist::Zilla compliant and tested (closes #33)
- added _dzil_shim() and _dzil_unshim() for Dist::Zilla distributions
- bumped prereq of TestAgainst plugin to 0.06
- added -T (--selftest), to prevent recursive testing loops when
testing this distribution
- POD updates
- stderr.bblog as well as cpanm build.log files are now aggregated into
the FAIL log reports (closes #61 and #62)
- added -T, --selftest, to avoid recursive testing of Test::BrewBuild
with the BBDEV_TESTING env var set
1.04 2016-04-01
- ensure FAIL files are created properly and test (closes #53)
- fix issue where --on wasn't setting up the environment correctly
(fixes #54)
- on Unix, we catch if an installed version is attempting to be
installed, instead of letting perlbrew handle it (closes #50)
- prereq 'version' added to correctly get version info from pre 5.10
(fixes #51)
- renamed --version|-v to --install|-i
- added --notest|-N, allows install/remove without running exec()
(closes #49)
- added --setup|-s, displays Windows and Unix test platform setup guide
- added --legacy|-L. Due to the fact ``cpanm'' requires 5.8.1, and we
use it in the default exec plugin, we won't install versions lower
than that, unless this flag is set
- FAIL logs are now stored in a temp directory, then moved to './bblog'
after completion, ensuring manifest.t tests don't fail (fixes #56)
- fixed bug where on Windows, if a module build fails, the revdeps
could still pass (fixes #57)
- POD updates
- created berrybrew PR dnmfarrell/brewberry#16 which added the --with
arg to exec
- bumped prereq of Test::BrewBuild::Plugin::TestAgainst to 0.05 due
to the Windows path issues
1.03_03 2016-03-29
- --revdep, auto-testing of CPAN down-river modules working very well
(work on #39)
- added t/dev-1xx tests, need perlbrew/berrybrew installed to run. Set
BBDEV_TESTING environment variable to run these tests (these tests
can take a very long time to process, and currently, may be
unreliable)
- added Archive::Extract prereq for BBDEV_TESTING
- fixed bug where --version wasn't doing anything
- fixed bug on Windows where instance_remove() was removing the
currently in-use perl instance
- if an invalid arg is sent in, we now print help screen and exit
(closes #45)
- fix bug in results(), where the regex wasn't capturing properly on
Windows
- added blurb in SYNOPSIS about the importance of having a central
cpanm (closes #44)
- corrected a whole bunch of inconsistencies between win/nix throughout
the dev test writing
- we now re-check brew_info after an install, so we don't exit
prematurely thinking no perls are installed, even after a --new 1
(closes #46)
- had to change the DATA section in the plugins to one cmd per line,
due to Windows not properly able to manage PATH env var on each exec
run. The actual command to perform the tests *must* be the last line
executed
- added BrewExec::using(), sends back the exact version we're using, so
remove_instance() does the right thing on Windows (fixes #47)
- disabled 'on' tests on Windows, as berrybrew doesn't have a --with
flag for 'exec'. We warn() if it's attempted
- added examples/config.txt, shows a working Windows %PATH% env setup
- began, albeit lightly, code cleanup (#36)
- set objectives for v1.05 (full win/nix compat, document pristine
testing environment, fix all outstanding non-wishlist issues and
at minimum, document irregularities of test failures beyond this
modules control)
1.03_02 2016-03-26
- FAIL strategy completed. If there's STDERR output, we write it all to
brewbuild_err.bblog. FAIL reports go into $perlver.bblog, and if
we're using a plugin that accepts a module, we put fail reports into
Tested::Module-$perlver.bblog, with the '::' replaced into '-'. The
results (PASS and FAIL) are now all displayed at the end of output
for easy identification (closes #43)
- fixed case-insensitivity bug with options. -R and -r now do the
respective Right Thing
- cpanm is now called with the -v flag on the --test-only portion of
the default plugin
1.03_01 2016-03-25
- if --args is passed in, brewbuild runs the number of times the list
has elements, and on each pass, the plugin receives the current arg
in the list
- --args executes --new and --remove only on first run through the loop
- plugins now get a copy of the parent log for cloning
- added --revdep, tests all CPAN reverse dependencies on the module
you're testing
- added prereq CPAN::ReverseDependencies for --revdep
- Plugin::Simple v0.06 now required (fixes #28, #30)
- documented $log as first param to a plugin (closes #35)
- cleaned up SYNOPSIS (closes #38)
- add tests for --args param for plugins (closes #32)
- we now croak if there's no 't/' dir, as there'd be nothing to test
(closes #34)
- removed -R flag, -r was interfering. We'll figure this out later
- cleaned up logging. By default 0 will display install/remove actions,
all other logging needs debug 5-7 (closes #37)
1.02 2016-03-24
- default plugin now uses cpanm for all installs, rendering the build
system in use irrelevant (should work for distzilla) (closes #29)
- added new --args arg to 'brewbuild', this is an aref that gets
dereferenced and the list passed into the 'brewbuild_exec()' sub
in each plugin
- fixed abstract issue (closes #27)
- removed t/11-results.t, as we don't return results anymore
- modified logging levels
1.01 2016-03-22
- set README to the binary, not the module... d'oh!
1.00 2016-03-22
- removed Plugin.pm, we now use Plugin::Simple for the plugin init work
- BUILD_REQUIRES Mock::Sub v1.06 for testing purposes, so we can test
without *brew installed (and if it is, we don't want to be managing
instances)
- finally got around to implementing a near full test suite (closes #4)
- removed version() from BrewCommands.pm, not needed
- bumped major version number now that we're tested and stable
0.06 2016-03-21
- added '--on' param, send in a list of perl versions to run against,
instead of the default all (closes #17)
- renamed 'count' param to 'new', kept change for backwards
compatibility reasons (closes #23)
- we can now load non-installed local plugins by specifying the full or
relative path name plus name of file to the --plugin param (closes
#24)
- all *brew command creation moved to Test::BrewBuild::BrewCommands.pm
(closes #19)
- new round of unit tests
- added debug logging to Test::BrewBuild::BrewCommands
- added warning in Makefile.PL and skip a bunch of tests if *brew isn't
installed
- 'brewbuild' command warns with a link to the respective *brew, and
exits if it isn't installed
0.05 2016-03-18
- we now recognize when Module::Build is being used, as opposed
to EUMM, and it works properly (fixes #12)
- *brew exec commands are now loaded via a plugin architecture. Set by
the '-p|--plugin' or an environment variable, plugin writing is
exceptionally simplistic
- renamed --reload arg to --remove
- fix undef issue in Plugin.pm if a plugin name can't be found
- --debug param now accepts a level param (0-7) to set the level (added
Logging::Simple as prereq)
- remove all prereqs from Makefile.PL that are already in core
(closes #15)
- add prereqs that aren't in core as of v5.8 into Makefile.PL (closes
#20)
- completely peppered the modules with various debugging level
statements
- added POD in Test::BrewBuild::Plugin
0.04 2016-03-16
- testing now properly works across all installed perls
- the exec file is now auto-generated with File::Temp, and unlinked
after use (closes #9)
- install now skips and warns (w/ debug) if a version it's trying to
install is already installed (closes #8)
- added POD for both the script and module (closes #5)
0.03 2016-03-01
- App::cpanminus now a prereq
- fixed issue in split in results() (fixes #6)
0.02 2016-02-23
- implemented Object Oriented approach
- added _test_file(), returns the test file brew will run (uses
File::Temp)
- added t/01-test_file.t
- added POD (closes #3)
- version arg now accepts multiple values (closes #1)
- fixed defined-or to be 5.8 compatible
0.01 2016-02-23
First version, released on an unsuspecting world.