"Charles B. Owen" <[EMAIL PROTECTED]> writes: >It's a known bug that gdb (I've tried up to 5.0) fails when debugging any >executable that includes pthreads under Irix 6.5 (fails with unknown signal >message and locks up). I tried mapping the various signals, but could not >get it to work. Has anyone got a fix in the works? Does anyone have any >ideas of how to debug g++ code on an Irix machine that uses threads short of >lots of print statements? Any help will be greatly appreciated. Here's the situation as I understand it. FWIW. The only means for debugging IRIX6.5 pthreads apps is to use a shared library named 'libspypt.so' (the debugger itself must do this) and a bunch of unique-to-irix calls into libspypt.so. This library interface is shipped with IRIX, so the IRIX folks ensure the two work together across IRIX changes -- while keeping the interface constant so the debugger executable itself is independent of which IRIX6.5 is in use. AFAIK, only IRIX debuggers (dbx, WorkShop) and a perhaps-unreleased version of TotalView (from Etnus) know the interface. And the demangler library code in dbx and WorkShop does not know about g++ demangling. The pthreads-debug interface is a bit complicated because of the, um, messy relations between per-thread and per-process actions. We've not documented the interface publicly, AFAIK. It's not a secret, it's just not documented publicly yet. It's not a simple extension of the /proc ioctl's... I do hope to contribute gdb patches on this some time this year. (of course with comments explaining the interface). (I'll be on vacation for 6 weeks starting July 11, so don't hope for any action soon...). Sorry. David B. Anderson [EMAIL PROTECTED] [EMAIL PROTECTED] http://reality.sgi.com/davea/
