[uClibc] buildroot, toolchain or root_fs?

Carl Miller chaz at energoncube.net
Wed May 5 15:16:55 UTC 2004


On Wed, May 05, 2004 at 02:58:04PM +0200, Henri van Riel wrote:
> 
> I've build gcc-3.3.2 now. I had 1 segmentation fault during the build.
> 
> The workaround for that seems to be to go into the directory where it
> failed and type `make <program-it-was-trying-to-compile>.o`. Chances
> are it will compile without a problem and then return to top-level and
> type `make` again. Compilation will continue where it left off...
> 
> I'm now trying to build buildroot but I had 2 segfault so far. I did
> the same trick as described above and it's still going. Obviously the
> segfaults don't have much to do with the gcc version.

That sounds like the infamous "gcc fatal sig11" bug.  Which isn't a
software bug in gcc.  It's a hardware bug.  It turns out that using
gcc to compile large, complex packages is a better memory tester than
most dedicated memory testers.  If you get unexplained segfaults in
gcc that appear and disappear at random given the same input files and
compilation flags, you should suspect that you have marginally flaky
memory on your box.  Try replacing it with higher-quality, faster-rated,
or better-cooled (or any combination thereof) RAM bricks and see if the
problem clears up.


                               -----Carl



More information about the uClibc mailing list