http://www.iki.fi/jhi/perl@xxxxxxxxx
http://www.iki.fi/jhi/perl@xxxxxxxxx
(or rsync -avz ftp.linux.activestate.com::perl-5.8.x perl-5.8.x)
(This just in after I already rolled the tarballs: Win32ers may
need to add dTHX; at the top of the Perl_doing_taint() in perl.c)
Changes since the last:
____________________________________________________________________________
[ 19891] By: jhi on 2003/06/30 09:39:29
Log: Integrate:
[ 19857]
Regen Changes.
[ 19858]
perlhack update, by Steve Grazzini
about macro support in gdb and gcc.
[ 19859]
Hash/Util.t and Encode/t/Aliases.t seem to be having
random failures. To make these easier to reproduce,
add a variable, PERL_HASH_SEED_DEBUG, to display the
hash seed. E.g. in Debian/x86 Linux 3.0 PERL_HASH_SEED
of 82972356 makes the first one to fail.
[ 19860]
Subject: Re: Change 19854: Bite the bullet and apply the hash
randomisation patch.
From: Tim Bunce <Tim.Bunce@xxxxxxxxx>
Date: Thu, 26 Jun 2003 10:53:22 +0100
Message-ID: <20030626095322.GE97463@xxxxxxxxxxxxxxxxxxxx>
[ 19861]
Do not obey PERL_HASH_SEED or PERL_HASH_SEED_DEBUG
if tainting-- but is this a good thing or a bad thing?
(At least it makes debugging lib/Hash/Util.t harder,
since it has, for no apparent good reason, -T: one must
make a copy of it without the -T.)
[ 19862]
Make doing_taint() always available (though not
part of the public API).
[ 19863]
Introduce (global) variable PL_earlytaint which
is set very early in main(), before perl_parse()
has been called and PL_tainting (or PL_taint_warn)
might have been set.
[ 19864]
Use the PL_earlytaint. (PL_earlytaint is a global,
not per-interp, since perl_construct() is not passed
the argc, argv, and therefore it can't set the per-interp
PL_tainting.)
[ 19865]
atoi() doesn't cut the mustard if the PERL_HASH_SEED
is larger than INT_MAX (atoi() returns -1 in that case).
[ 19866]
Some warnings about the (im)proper uses of the hash randomisation.
[ 19867]
The two-for-loops is no more a valid way to walk through
a hash (this was the reason the Hash/Util.t intermittently
failed, the two-loop didn't find all the SVs of the HV).
[ 19868]
Integrate mainline
[ 19869]
Fix test count, by Abe Timmerman.
[ 19870]
Two debugging patches.
The first allows to hold symbolic switches in $^D
and more generally fixes assignment to $^D. The
second one improves the information given by -Dl.
Subject: [PATCH] allow $^D = "flags"
From: Dave Mitchell <davem@xxxxxxxxxxx>
Date: Fri, 27 Jun 2003 22:26:24 +0100
Message-ID: <20030627212624.GB12887@xxxxxxxxxxx>
Subject: [PATCH] make -Dl show more scope info
From: Dave Mitchell <davem@xxxxxxxxxxx>
Date: Fri, 27 Jun 2003 23:00:36 +0100
Message-ID: <20030627220036.GC12887@xxxxxxxxxxx>
[ 19871]
Subject: [Encode] pre-1.97 patches
From: Dan Kogai <dankogai@xxxxxxxxx>
Date: Sat, 28 Jun 2003 01:20:59 +0900
Message-Id: <56D5BFEE-A8BB-11D7-9092-000393AE4244@xxxxxxxxx>
[ 19872]
Some clarification about the current semantics of CHECK and
INIT blocks. See bug [perl #22826].
[ 19873]
Using $1 without testing success of the regexp, bad.
[ 19874]
Retract #19867; the bug was really much simpler:
the < max must be <= max instead.
[ 19875]
Duh.
[ 19876]
Subject: Re: your malloc patches
From: Ilya Zakharevich <ilya@xxxxxxxxxxxxxxxxx>
Date: Fri, 27 Jun 2003 06:54:06 -0700
Message-ID: <20030627135406.GA8914@xxxxxxxxxxxxxxxxx>
More malloc patches: now they seem to work even in Tru64.
[ 19877]
The #19842 is no more needed thanks to #19876,
and the #19842 was wrong anyway (it affected
only the threaded case.)
[ 19878]
Move the PL_earlytaint initialization to the PERL_SYS_INIT()
as per suggestion from Sarathy.
[ 19879]
Another spot where a zero $test{$max} can make things go boom.
[ 19880]
argc, argv.
[ 19881]
More coffee...
[ 19882]
Perl_doing_taint must be public, for programs that embed perl
[ 19883]
More on the macro debugging and expansion.
[ 19884]
The joy of $0. Undoing the #16399 makes Andreas'
tests (see [perl #22811]) pass (yes, padding with space instead
of nul makes no sense, but that seems to work, maybe Linux does
some deep magic in ps(1)?); moving the PL_origalen computation
earlier makes also the threaded-first case fully pass.
But in general modifying the argv[] is very non-portable.
(e.g. in Tru64 it seems to be limited to the size of the
original argv[0] since the argv[] are not contiguous?)
Everybody should just have setproctitle().
[ 19885]
Fix a faulty alias.
[ 19886]
Misc Pod Nits.
[ 19887]
$0 test tweaks from Andreas.
[ 19888]
$0 doc tweakage.
[ 19889]
The 'contiguous' test for argv[], envp[] was bogus
since those need not be in memory end-to-end, e.g.
in Tru64 they are aligned by eight. Loosen the test
so that 'contiguousness' is fulfilled if the elements
are within PTRSIZE alignment. This makes Tru64 to pass
the join.t, too.
[ 19890]
int is not UV.
Branch: maint-5.8/perl
!> (integrate 39 files)
____________________________________________________________________________
[ 19857] By: jhi on 2003/06/26 08:29:59
Log: Regen Changes.
Branch: maint-5.8/perl
! Changes
____________________________________________________________________________
[ 19856] By: jhi on 2003/06/26 05:55:24
Log: Update Changes.
Branch: maint-5.8/perl
! Changes patchlevel.h
____________________________________________________________________________
--
Jarkko Hietaniemi <jhi@xxxxxx> http://www.iki.fi/jhi/ "There is this special
biologist word we use for 'stable'. It is 'dead'." -- Jack Cohen
|
