Description of problem: glibc errata busted gdb Version-Release number of selected component (if applicable): 2.3.2-4.80 How reproducible: Try to debug Mozilla. Steps to Reproduce: 1. Try to start up Mozilla in gdb. Actual results: gdb will start spewing errors like: Program received signal SIG32, Real-time event 32. Expected results: Program should be debugged as normal. Additional info: This happened when the glibc update was released. Lots of people are reporting this. We need to unscrew it.
Can you please try ftp://people.redhat.com/jakub/glibc/errata/8.0/ ?
In case anyone finds it useful, I filed a detailed report about this bug as a gdb bug, number 1154. You can see that here: http://sources.redhat.com/cgi-bin/gnatsweb.pl I wasn't debugging mozilla, but a different large, multithreaded program. Same results though: gdb doesn't work at all for my program.
FYI in the errata there is no NPTL, just two copies of linuxthreads (floating stacks and non-floating-stacks). Because of a bug in the kernel and bugfix in glibc only non-floating-stacks linuxthreads are used in 2.3.2-4.80, so in 2.3.2-4.80.1 I've backed out the glibc bugfix so that floating-stacks libpthread can be used unless LD_ASSUME_KERNEL=2.2.5 or kernel too old.
I tried the 2.3.2-4.80.1 packages from ftp://people.redhat.com/jakub/glibc/errata/8.0/ Didn't help. Gdb still refuses to recognize threaded programs.
Oops, sorry. Actually more work was needed to make /lib/i686 libs to be used even when the kernel supplies broken AT_PLATFORM aux vector value. Can you try: ftp://people.redhat.com/jakub/glibc/errata/8.0/*4.80.3* ?
I'm no longer seeing the SIG32 problems with *4.80.3 packages
*4.80.3 fixed my gdb problems as well.
Should be fixed in the last errata.