This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Seg faults/Internal compiler errors when using Linux swapfil
- To: egcs-bugs at cygnus dot com (IPM Return requested)
- Subject: Seg faults/Internal compiler errors when using Linux swapfil
- From: John Linebarger <jmlineb at sandia dot gov>
- Date: 14 Jul 1998 15:16:11 -0600
- Alternate-Recipient: Allowed
- Conversion: Allowed
- Disclose-Recipients: Prohibited
- Original-Encoded-Information-Types: IA5-Text
- X400-Content-Type: P2-1988 ( 22 )
- X400-MTS-Identifier: [/c=US/admd= /prmd=USDOE/; 0640735ABCA9B030-mtaSNL]
- X400-Originator: jmlineb@sandia.gov
- X400-Received: by mta mtaSNL in /c=US/admd= /prmd=USDOE/; Relayed; 14 Jul 1998 15:16:11 -0600
- X400-Received: by /c=US/admd= /prmd=USDOE/; Relayed; 14 Jul 1998 15:16:11 -0600
- X400-Recipients: non-disclosure;
You're gonna hate me for the lack of specificity here, but I've been
using egcs as my compiler for months now, and I've noticed a problem.
At least under Slackware Linux, kernel 2.0.34, with a 64M swap file,
running on a 233MHz Pentium MMX PC with 64M of main memory and lots of
EIDE hard drive space, here's what I consistently notice:
1. Seg faults (sig 11) during compilation.
2. "Internal compiler error" messages
3. "Attempt to remove invalid swap file page" messages (rare)
In *each* case, I simply retype "make," and the compiler just picks up
where it left off. I often have to do this several times in trying to
compile large packages or the Linux kernel.
The swap file message gave me a clue, so I added more memory to my
Linux box (I'm running at 128M now), and removed all swap partitions.
I noticed that the incidence of the above errors sharply decreased,
but did not go away entirely.
So I conclude that egcs has difficulty in using Intel Linux swap
space; it is especially bad when running in an Xterm window in
X-Windows, where less space is available. It is better running in raw
console mode.
However, at work where I have a 333 MHz Pentium-II with 256 MB of
SDRAM memory and a similar Linux software configuration and no swap
file, I get even fewer of the errors I describe above.
Is this an egcs bug? A Linux swap file bug? Are other people
reporting similar impressions and experiences?
Thanks!
John M. Linebarger
Sandia National Laboratories