Re: Memory attribute on movstrsi for i386

[Toon Moene <toon at moene dot indiv dot nluug dot nl>]

Richard Kenner wrote:
> 
> It's always a bad idea for an MD file to emit an explicit MEM.  Here, it does
> so in numerous places, but doesn't attempt to copy the attributes from the
> original memory reference.  This is an efficienty issue when the alias set
> isn't copied, but a correctness issue when RTX_UNCHANGING_P isn't clobbered,
> and the latter is how I caught this.
> 
> This fix is somewhat of a kludge, since the patterns in question need to be
> reworked to not have a MEM in the template, but there are quite a few pattern
> involved and I'm not in a good enough debugging position on that target to
> try such a massive change.  It should be done, though.

Hmmm, I can't bootstrap on i686-pc-linux-gnu with this change:

stage1/xgcc -Bstage1/ -B/usr/tmp/i686-pc-linux-gnu/bin/ -c  -DIN_GCC   
-g -O2 -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes
-Wtraditional -pedantic -Wno-long-long  -DHAVE_CONFIG_H
-DGENERATOR_FILE    -I. -I. -I../../gcc/gcc -I../../gcc/gcc/.
-I../../gcc/gcc/config -I../../gcc/gcc/../include ../../gcc/gcc/rtl.c -o
rtl.o
../../gcc/gcc/rtl.c: In function `read_rtx':
../../gcc/gcc/rtl.c:1277: Internal error: Segmentation fault
Please submit a full bug report, with preprocessed source if
appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.

So I patch -R'd it and rebuilt (using --disable-checking and
--enable-languages=c,f77 for configure in both cases).

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)