.
Today, I will mostly be crashing gcc. Repeatedly. - Grin with cat attached — LiveJournal
Previous Entry Next Entry
Today, I will mostly be crashing gcc. Repeatedly. Jul. 19th, 2004 04:03 pm
PHP 5 compiles fine on one bytemark box and dies 'orribly on the other. Seems more than likely that the gcc v3.3.4-2 on heifong is the culprit; tenrei has 2.95.4-14 and seems fine.
Can't for the life of me remember what package on heifong demanded the gcc "upgrade".


/bin/sh /home/wechsler/tarbs/php-5.0.0/libtool --silent --preserve-dup-deps --mode=compile gcc -DSUPPORT_UTF8 -DLINK_SIZE=2 -DPOSIX_MALLOC_THRESHOLD=10 -I/home/wechsler/tarbs/php-5.0.0/ext/pcre/pcrelib -Iext/pcre/ -I/home/wechsler/tarbs/php-5.0.0/ext/pcre/ -DPHP_ATOM_INC -I/home/wechsler/tarbs/php-5.0.0/include -I/home/wechsler/tarbs/php-5.0.0/main -I/home/wechsler/tarbs/php-5.0.0 -I/home/wechsler/tarbs/php-5.0.0/Zend -I/usr/include/libxml2 -I/usr/include/mysql -I/usr/local/include -I/home/wechsler/tarbs/php-5.0.0/TSRM -g -O2 -prefer-pic -c /home/wechsler/tarbs/php-5.0.0/ext/pcre/pcrelib/pcre.c -o ext/pcre/pcrelib/pcre.lo
gcc: Internal error: Killed (program cc1)
Please submit a full bug report.
See <url:http://gcc.gnu.org/bugs.html> for instructions.
For Debian GNU/Linux specific bug reporting instructions,
see <url:file:///usr/share/doc/gcc-3.3/readme.bugs>.

make: *** [ext/pcre/pcrelib/pcre.lo] Error 1

From: deborah_c
Date: July 19th, 2004 - 03:33 pm (Link)
This is at least one step better than having it just generate wrong code (which I've also known, although *mostly* in gcc versions where the back end is not a mainstream one). However, you have my sympathies as a mostly reformed gcc hacker.
(no subject) - xraycb
From: wechsler
Date: July 19th, 2004 - 04:45 pm (Link)
On Linux?

Debian-unstable running under user-mode linux. Heifong should have at least as much swap as tenrei, and has twice the RAM.

Dmesg is *full* of out-of-memory errors for various services :/
From: conflux
Date: July 19th, 2004 - 10:09 pm (Link)
But is it running more memory intensive processes? I've also found that various make utils that we have used at work can leave other bits of the iterative build process running if a gcc operation dies like this.
(no subject) - (Anonymous)
From: wechsler
Date: July 20th, 2004 - 07:02 am (Link)
Nonsense

But near-correct nonsense. Apache had gone rogue and was using all the memory it could get its paws on. Killing it (and incidentally updating to the latest 1.3.x) solved the problem.