Revision history for Perl extension File::Tail.
0.01 Fri Nov 28 09:44:06 1997
- original version; created by h2xs 1.18
0.5 Fri Nov 28 12:25:13 MET 1997
- original version; created by matija.grabnar@arnes.si
0.6 Thu Dec 4 14:16:59 MET 1997
- assigning error values to $! didn't work. Replaced with croaks.
- passing only the filename didn't work.
0.7 Mon Jun 1 17:23:23 MET 1998
- implemented tied interface
- implemented non-blocking read (against my better judgement)
- implemented more flexible error handling a-la Net::Telnet
- implemented ignore_nonexistat
Release of 0.7 was delayed because I wanted to see if Tied
Handle handling will include select in 5.005. After two people
sent me their tied interfaces, I figured maybe it was time to
release it before implementing select :-)
0.8 Wed Oct 14 19:29:30 MET DST 1998
implemented wantarray handling in READLINE
implemented some tests for make test
documented the tied interface in synopsis (Duh!)
- implemented tail -n functionality
0.81 Thu Oct 15 07:54:36 MET DST 1998
There was a bug in tail -n part, and the tail -n tests were missing
from the distribution
0.85 Wed Oct 28 15:44:47 MET 1998
The introduction of tail -n broke the handling of files where
a line might be written incompletely (e.g. without the final \n). OOps!
0.90 Tue Nov 10 11:02:35 MET 1998
fix for reset_tail (it couldn't be set to 0) from
Chris_Heller/PeopleSoft@peoplesoft.com
fixed some -> references which complained in Perl5.4 but not in 5.5
moved debugging statement to a separate routine, logit
logit now outputs the name of the calling routine and the name of the
file the object refers to.
select functionality added.
0.91 Tue Feb 2 17:14:27 MET 1999
another ambigous reference (fixed thanks to Peter Wirdemo <edtpewo@ratatosk.aom.ericsson.se>)
nasty select bug discovered by "Weeber, Steve" <steve.weeber@qwest.net>
select can now also be called as File::Tail::select, as long as there
is at least one tail object given.
0.92 Thu Oct 21 11:28:57 MET DST 1999
Blair Zajac sent a patch to fix some problems when returning an empty
string when the request was for an array.
Gerben Wierda <Gerben_Wierda@RnA.nl> sent a patch that should fix
problems with files that don't get updated for a while (a file could
accidentaly be read the whole way through again under some
circumstances)
0.94 Wed Nov 3 12:46:29 MET 1999
Some changes in the way internal data caching is done - primarily
support for cases where you shouldn't keep all the pending lines in
memory.
logit calls are now stripped by default before the module is installed.
If you want to patch File::Tail now, you must edit Tail.pm.debug,
the do a perl Makefile.PL which will create a Tail.pm
Select returned ALL handles instead of just the ones with input ready.
0.95 Thu Dec 9 10:25:47 MET 1999
I couldn't get Pause to recognise that 0.94 was newer than 0.91 for
some reason. I added Tail.pm back into the manifest to see if that
will help.
0.96 Mon May 29 09:33:27 MET DST 2000
added binmode to improve running on Microsoft platforms
added a check in DESTROY to avoid a warning in Perl 5.6
fixed prerequisites in Makefile.PL
0.97 Persistent "go wild after log rotation, but only sometimes" bug
perhaps eliminated. Thanks elfchief@lupine.org for making a script
that triggered the previously interminent problem every time.
0.98 David Pfitzner release - David send me 6 patches (some for bugs,
others for inefficiencies, or lack of clarity in documents)
0.99 "Lyle D. Brooks" <brooks@deseret.com> reported there were instances
where interval was set to a negative value
Joe Smith <Joe.Smith@wcom.com> noted that one shouldn't declare
@ISA=qw(Autoloader) unless one is prepared to deal with it correctly.
Peter Allen <pgmppa@ibi.com> undef warnings when starting up on a
nonexistant file
Alain Fauconnet <alain@ait.ac.th> and Benjamin Zwittnig <beni@arnes.si>
provided test cases which helped me chase down another two cases where
File::Tail would spontaneously read too much.
0.99.2 Some operating systems sometimes return 0 for sysread on busy files.
This should now be handled correctly.
nowait can now be specified at object creation.
A callback is called on every file rotation, which can be used for
files that change their names with time.
0.99.3 Firefox crashed while I attempted upload to PAUSE, so I had to make
a new version just for that. Life is ridiculous.
1.0 Actually just two minor bug fixes (one of them in a test), but I no
longer see a point in not having a 1.0 version.
1.1 Lee Duncan drew my attention to Stephan Muller's fixes for Windows compatibility
Changed the use of the system's mv command to using File::Copy in the
tests. (Steffen Mueller)
Added machine-readable license statement to Makefile.PL and thus
META.yml (Steffen Mueller)
The sixth test in 10/open.t is skipped on win32 because you can't just
move files around that are opened. (Steffen Mueller)
Due to using sysread and friends, there were newline problems on win32.
That should be fixed now. (Steffen Mueller)
1.2 Break the infinite loop that can result when the average length of lines
causes the attempt to fill the tail buffer to fill with the exact same
or even smaller number of lines.
1.3 Fix for a stupid bug in 1.2 (GFILATOV, Slaven_Rezic)
Added a warning for use of debug in a non-debug version of File::Tail
Shows a warning when maxbuf is set to a too-small value
Invoking name_changes callback changes the value of input attribute (sottile@ix.netcom.com)
When deciding to reopen the file, check if the inode matches (that would mean it has not
been ranamed)