[Python-checkins] python/dist/src/Modules _sre.c,2.87,2.88 sre_constants.h,2.13,2.14

[Andrew MacIntyre]

On Fri, 18 Apr 2003, Andrew MacIntyre wrote:

> On Mon, 14 Apr 2003 gvanrossum@users.sourceforge.net wrote:
>
> > Update of /cvsroot/python/python/dist/src/Modules
> > In directory sc8-pr-cvs1:/tmp/cvs-serv12812/Modules
> >
> > Modified Files:
> > 	_sre.c sre_constants.h
> > Log Message:
> > SF patch #720991 by Gary Herron:
> > A small fix for bug #545855 and Greg Chapman's
> > addition of op code SRE_OP_MIN_REPEAT_ONE for
> > eliminating recursion on simple uses of pattern '*?' on a
> > long string.
>
> This patch is causing a bus error on FreeBSD 4.4 (hope to check 4.7 later
> this weekend).  The output of a verbose test_sre run is
>
> test_sre
> Running tests on character literals
> Running tests on sre.search and sre.match
> bus error core dumped
>
> Compiler is gcc 2.95.3, with -O3.

Dropping the optimisation to -O2 for _sre.c resolves the problem.  I need
to check my 4.7 system, which has gcc 2.95.4, to see whether it also has
the same problem.  I'll also try the gcc 2.95.2 compiler on OS/2.

I recompiled the whole of Python with -O2, and on the 4.4 system (a
Celeron 300A SMP box) the pystone rating goes from ~4550 (-O3) to ~4565
(-O2).

I'd appreciate suggestions about formulating a patch, as I know
essentially zip about autoconf...

> Backing the patch out (effectively reverting to 2.3a2) results in the
> attached log file, which looks like some sort of Unicode issue, where
> 2.3a2 builds and tests fine on the same box.

Forget that bit of blather...

--
Andrew I MacIntyre                     "These thoughts are mine alone..."
E-mail: andymac@bullseye.apana.org.au  (pref) | Snail: PO Box 370
        andymac@pcug.org.au             (alt) |        Belconnen  ACT  2616
Web:    http://www.andymac.org/               |        Australia