Re: [Valgrind-users] Callgrind: threads.c:396 (push_exec_state): Assertion

[Maxim Y. <max...@gm...>]

Sorry, forgot to forward it to the list.

On 22/10/06, Maxim Yegorushkin <max...@gm...> wrote:
> On 22/10/06, Josef Weidendorfer <Jos...@gm...> wrote:
> > On Sunday 22 October 2006 16:24, Maxim Yegorushkin wrote:
> > > My program uses signals heavily (sends itself).
> >
> > Actually, a good hint.
> >
> > > BB# 343925
> > >
> > > Callgrind: threads.c:396 (push_exec_state): Assertion '(sp > 0) && (sp
> > > < 10)' failed.
> >
> > Wow.
> > This is one of the few places where callgrind has a fixed limit.
> > You must be running more than 9 signal handlers at the same time
> > (in one thread).
> >
> > Could this be the case?
>
> I'm using 16 realtime signals [SIGRTMIN, SIGRTMIN + 15]. All the
> signal handler does is writing a byte to a pipe.
>
> > Can you try to increase MAX_SIGHANDLERS in callgrind/global.h,
> > e.g. to 30, and test your program again?
> >
> > Hmmm... if you allow for reentrency into a signal handler, 30 could
> > be not enough.
> >
> > Can you tell us if this helps?
>
> I took the easy path by making my program use 8 signals, rather than
> 16 (the program is a test, so it's easily tunable). Now I get another
> assertion. I can recompile valgrind to investigate further if you
> like.
>
> [max@k-pax demux]$ valgrind -v --tool=callgrind bin/test > log
> ==3466== Callgrind, a call-graph generating cache profiler.
> ==3466== Copyright (C) 2002-2006, and GNU GPL'd, by Josef Weidendorfer et al.
> ==3466== Using LibVEX rev 1658, a library for dynamic binary translation.
> ==3466== Copyright (C) 2004-2006, and GNU GPL'd, by OpenWorks LLP.
> ==3466== Using valgrind-3.2.1, a dynamic binary instrumentation framework.
> ==3466== Copyright (C) 2000-2006, and GNU GPL'd, by Julian Seward et al.
> ==3466==
> --3466-- Command line
> --3466--    bin/test
> --3466-- Startup, with flags:
> --3466--    -v
> --3466--    --tool=callgrind
> --3466-- Contents of /proc/version:
> --3466--   Linux version 2.6.17-1.2187_FC5
> (bre...@hs...) (gcc version 4.1.1 20060525
> (Red Hat 4.1.1-1)) #1 Mon Sep 11 01:17:06 EDT 2006
> --3466-- Arch and hwcaps: X86, x86-sse1-sse2
> --3466-- Valgrind library directory: /usr/local/lib/valgrind
> ==3466== For interactive control, run 'callgrind_control -h'.
> --3466-- Reading syms from /lib/ld-2.4.so (0x526000)
> --3466-- Reading syms from /home/max/src/demux/bin/test (0x8048000)
> --3466-- Reading syms from /usr/local/lib/valgrind/x86-linux/callgrind
> (0x38000000)
> --3466--    object doesn't have a dynamic symbol table
> --3466-- Code check found runtime_resolve: ld-2.4.so
> +0x12B60=0x538B60, length 24
> --3466-- Reading syms from
> /usr/local/lib/valgrind/x86-linux/vgpreload_core.so (0x4001000)
> --3466-- Reading syms from /lib/librt-2.4.so (0xDA4000)
> --3466-- Reading syms from /usr/lib/libstdc++.so.6.0.8 (0x3E7000)
> --3466--    object doesn't have a symbol table
> --3466-- Reading syms from /lib/libm-2.4.so (0xDB3000)
> --3466-- Reading syms from /lib/libgcc_s-4.1.1-20060525.so.1 (0x3D9000)
> --3466--    object doesn't have a symbol table
> --3466-- Reading syms from /lib/libpthread-2.4.so (0x3C3000)
> --3466-- Reading syms from /lib/libc-2.4.so (0x101000)
> --3466-- Symbol match: found runtime_resolve: ld-2.4.so +0x538B60=0x538B60
> demux::engine_base::signal_handler(int)
> BB# 337249
>
> Callgrind: threads.c:273 (vgCallgrind_post_signal): Assertion
> '*(vgCallgrind_current_fn_stack.top) == 0' failed.
> ==3466==    at 0x38019C11: report_and_quit (m_libcassert.c:136)
> ==3466==    by 0x38019F3B: vgPlain_assert_fail (m_libcassert.c:200)
> ==3466==    by 0x3801795D: vgCallgrind_post_signal (threads.c:273)
> ==3466==    by 0x3805ACC7: vgPlain_sigframe_destroy (sigframe-x86-linux.c:694)
> ==3466==    by 0x3805D89B: vgSysWrap_x86_linux_sys_sigreturn_before
> (syswrap-x86-linux.c:954)
> ==3466==    by 0x3804D75F: vgPlain_client_syscall (syswrap-main.c:719)
> ==3466==    by 0x38039E20: vgPlain_scheduler (scheduler.c:721)
> ==3466==    by 0x38058D53: run_a_thread_NORETURN (syswrap-linux.c:87)
>
> sched status:
>   running_tid=1
>
> Thread 1: status = VgTs_Runnable
> ==3466==    at 0x3CDD93: __write_nocancel (in /lib/libpthread-2.4.so)
> ==3466==    by 0x3CF0C7: (within /lib/libpthread-2.4.so)
> ==3466==    by 0x3CF0C7: (within /lib/libpthread-2.4.so)
> ==3466==    by 0x3CF0C7: (within /lib/libpthread-2.4.so)
> ==3466==    by 0x804AEA4: demux::engine_base::mask(int, __sigset_t
> const*, __sigset_t*) (demux.h:420)
> ==3466==    by 0x804B4B5: demux::engine_base::mask_guard::unblock()
> (demux.h:303)
> ==3466==    by 0x804BFB6: demux::epoll_::do_readiness(unsigned,
> demux::engine_base::mask_guard*, long long) (demux.h:642)
> ==3466==    by 0x804C347: demux::engine_<demux::epoll_>::loop() (demux.h:574)
> ==3466==    by 0x8049352: loop(unsigned) (test.cc:133)
> ==3466==    by 0x804948A: main (test.cc:150)
>
> Note: see also the FAQ.txt in the source distribution.
> It contains workarounds to several common problems.
>
> If that doesn't help, please report this bug to: www.valgrind.org
>
> In the bug report, send all the above text, the valgrind
> version, and what Linux distro you are using.  Thanks.
>