Red Hat Bugzilla – Bug 65276
Continual segfaults when compiling various programs
Last modified: 2007-04-18 12:42:41 EDT
From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 5.0; Windows 98; DigExt)
Description of problem:
After upgrading to RH7.3 I get get continual segfaults while compiling programs.
Does not seem to depend on the program - after segfaulting retyping make
Problem did not occur under RH7.2
Example programs gnome2, kernel,evolution
No useful information is produced at the terminal
I have tried;
re-installing glibc rpm,s
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Try to compile any reasonably sized program
2. Make stops with segmentation fault
3. Typing make again carries on compilation
Machine details K6/2 500 with 128MB ram
cpu often shows 100% usage
The following bugs all seem to relate to the same problem
As regards my particular bug exactly the same system with roswell did not give
I have noticed that the problem does get worse after a sustained period of
compiling, so I wonder if there is a memory leak in gcc, make,glibc or the
If it is any help it seems very similar to the problems I had with RH7.0 and
the original gcc2.96 which were fixed by using -O5 -mcpu=i586 -march=i586 (as
suggested by teg I think on slashdot)
so I am wondering if some default compiler flags were changed for 7.3
I assume the packages on RH7.3 were compiled with the compiler so would it be
an idea to post what options were used?
The current gcc + binutils fails to to compile the current kernel at link stage.
Problem massively reduced after compiling and running kernel 2.4.19-test10
I've repeatedly reproduced it when trying to compile Apache 2.0.36 from source
and PHP 4.2.1 from source.
The apache was compiled with the following configure line:
./configure --prefix=/www --enable-module=so
The php was compiled with:
./configure --with-mysql --enable-calendar --with-axps=/www/bin/axps
I managed to compile the apache via numerous retries, but I cannot trust the
resulting binary to be stable in a production system.
First one sounds like out of memory (new gcc's use a lot more ram).
Second reporter is clearly hardware