Re: tcl-related segfault in NAMD

From: Jérôme Hénin (jerome.henin_at_ibpc.fr)
Date: Wed Mar 06 2013 - 09:33:07 CST

Hi FX,

The backtrace hints at a possible problem with charm++ (in CsdScheduler / CsdScheduleForever). You didn't specify which version of charm++ you use, you might want to try a recent development version and if this segfault persists, report it to charm++ developers.

Note that for NAMD you shouldn't need to compile any fortran interface.

Cheers,
Jérôme

----- Original Message -----
> Hi all,
>
> After patching NAMD (the Langevin piston code, see my earlier email
> today), I have had to compile my own binaries based on the NAMD-2.9
> code, on my 64-bit Intel Mac OS 10.8.2 machine. I have successfully
> compiled FFTW (no problem), CHARM++ (took a lot of time to figure
> out, but it compiled okay and test programs ran fine), and TCL is
> already installed on the machine.
>
> So, I compiled NAMD with this configuration: "./config
> MacOSX-x86_64-g++ --charm-arch net-darwin-x86_64-gfortran-gcc
> --with-debug"
> Problem is, I now get a segfault at the beginning of the run. The
> backtrace, obtained from gdb, is the following:
>
> > Program received signal EXC_BAD_ACCESS, Could not access memory.
> > Reason: KERN_INVALID_ADDRESS at address: 0x00007fff02f47fb8
> > 0x00000001008d2660 in setJcontext () at stl_algo.h:2831
> > 2831 std::__final_insertion_sort(__first, __last);
> > (gdb) bt
> > #0 0x00000001008d2660 in setJcontext () at stl_algo.h:2831
> > #1 0x00000001008d26d1 in swapJcontext () at stl_algo.h:2831
> > #2 0x000000010095243b in CsdScheduleForever () at
> > ckreduction.h:700
> > #3 0x00000001009524cd in CsdScheduler () at ckreduction.h:700
> > #4 0x000000010006c3f6 in BackEnd::suspend () at BackEnd.C:234
> > #5 0x000000010080aca1 in ScriptTcl::suspend (this=0x10115a5c0) at
> > ScriptTcl.C:63
> > #6 0x0000000100811448 in ScriptTcl::initcheck (this=0x10115a5c0)
> > at ScriptTcl.C:89
> > #7 0x000000010080c7ed in ScriptTcl::Tcl_minimize
> > (clientData=0x10115a5c0, interp=0x101860a10, argc=2,
> > argv=0x10186e450) at ScriptTcl.C:466
> > #8 0x0000000100f5a59d in TclInvokeStringCommand ()
> > #9 0x0000000100f5b9cf in Tcl_ListMathFuncs ()
> > #10 0x0000000100f5ca7a in Tcl_EvalEx ()
> > #11 0x0000000100f5c444 in Tcl_EvalEx ()
> > #12 0x0000000100fb8e85 in Tcl_FSEvalFileEx ()
> > #13 0x0000000100fb7cbb in Tcl_EvalFile ()
> > #14 0x000000010080af25 in ScriptTcl::load (this=0x10115a5c0,
> > scriptFile=0x7fff5fbffbef "md.conf") at ScriptTcl.C:1139
> > #15 0x0000000100066f4c in after_backend_init (argc=2,
> > argv=0x7fff5fbffa98) at mainfunc.C:147
> > #16 0x00000001000666a1 in main (argc=2, argv=0x7fff5fbffa98) at
> > mainfunc.C:44
>
>
>
> Does anyone have a clue what is happening? How can I debug this
> further? Are there special tricks to get NAMD to compile correctly
> on Mac OS X?
>
> Thanks in advance,
> François-Xavier
>
>
>
> PS: I am compiling with Mac OS's native g++ compiler, which
> identifies itself as "gcc version 4.2.1 (Based on Apple Inc. build
> 5658) (LLVM build 2336.11.00)".
>
>

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:21:00 CST