Front page | perl.perl5.porters |
Postings from November 2003
On Tue, Nov 18, 2003 at 01:55:41PM +0100, Rafael Garcia-Suarez wrote:
> Jeff 'japhy' Pinyan wrote:
> >
> > This only happens with lexicals.
> >
> > my ($A, $B);
> > $B = ($A = '' . foo()) . "y";
> > $D = ($C = '' . foo()) . "y";
> > print "$A,$B\n"; # xy,xy
> > print "$C,$D\n"; # x,xy
> >
> > sub foo { "x" }
>
> Ouch.
> I can reproduce it with 5.6.1, but not with 5.5.3.
> Looks like an optree optimization problem.
Both the inner and outer OP_CONCATs have the TEMP target optimised away;
the inner one uses the lexical as the target instead, and the outer one is
STACKED, so it uses the previous result as the target (which happens to be
the lexical).
> Andreas, could you try this below with binsearchaperl ?
>
> #!perl
> sub foo { "x" }
> my ($A, $B);
> $B = ($A = '' . foo()) . "y";
> print "$A,$B" eq "x,xy" ? "ok\n" : "not ok\n";
> __END__
Looks like 3612 judging by these excerpts from the log:
____________________________________________________________________________
[ 4749] By: gsar on 2000/01/02 21:37:29
Log: disable optimization in change#3612 for join() and quotemeta()--this
removes all the gross hacks for the special cases in that change; fix
pp_concat() for when TARG == arg (modified version of patch suggested
by Ilya Zakharevich)
Branch: perl
! op.c opcode.h opcode.pl pp_hot.c sv.c t/op/lex_assign.t
____________________________________________________________________________
[ 4543] By: gsar on 1999/11/10 01:52:22
Log: remove dead branch/infinite looper in change#3612
Branch: perl
! op.c
____________________________________________________________________________
[ 4415] By: gsar on 1999/10/20 00:52:34
Log: disable optimizing troublesome ops in change#3612
(from Ilya Zakharevich)
Branch: perl
! Makefile.SH opcode.h opcode.pl
____________________________________________________________________________
[ 4087] By: gsar on 1999/09/06 18:06:06
Log: change#3612 is buggy when quotemeta argument matches target
(hope this is the last of the optimized-OP_SASSIGN bugs)
From: Ilya Zakharevich <ilya@math.ohio-state.edu>
Date: Sun, 5 Sep 1999 06:07:42 -0400 (EDT)
Message-Id: <199909051007.GAA06423@monk.mps.ohio-state.edu>
Subject: Re: [BUG: quotemeta]
Branch: perl
! Changes op.c t/op/lex_assign.t
____________________________________________________________________________
[ 3854] By: gsar on 1999/08/01 20:41:53
Log: fixes from Stephen McCamant that address bugs in change#3612
(the optimization shouldn't be enabled in expressions where
the variable is introduced), and fix Deparse to grok the
optimization
Date: Thu, 29 Jul 1999 21:21:49 -0500 (CDT)
Message-ID: <14241.3133.979257.953396@alias-2.pr.mcs.net>
Subject: [PATCH _58] Set OPpTARGET_MY more consistently
--
Date: Thu, 29 Jul 1999 22:31:16 -0500 (CDT)
Message-ID: <14241.7300.181386.763503@alias-2.pr.mcs.net>
Subject: [PATCH _58] Disable TARGET_MY-ization on variable introduction
--
Date: Fri, 30 Jul 1999 22:25:27 -0500 (CDT)
Message-Id: <199907310326.VAA24376@localhost.frii.com>
Subject: [PATCH _58, long] B::Deparse (was Re: New warning 'Useless use of...')
Branch: perl
! ext/B/B/Deparse.pm op.c
____________________________________________________________________________
[ 3664] By: gsar on 1999/07/11 19:11:07
Log: change#3612 was buggy and failed to build Tk; applied Ilya's
remedy and related tests via private mail
Branch: perl
! op.c t/op/lex_assign.t
____________________________________________________________________________
[ 3612] By: gsar on 1999/07/06 10:17:52
Log: From: Ilya Zakharevich <ilya@math.ohio-state.edu>
Date: Sat, 12 Jun 1999 04:49:09 -0400 (EDT)
Message-Id: <199906120849.EAA26986@monk.mps.ohio-state.edu>
Subject: [PATCH 5.005_57] Optimize away OP_SASSIGN
Branch: perl
! op.c op.h opcode.h opcode.pl pp.h pp.sym pp_proto.h
--
A walk of a thousand miles begins with a single step...
then continues for another 1,999,999 or so.
Thread Previous
|
Thread Next