On 27 Apr 2008, at 11:48, Martin Costabel wrote: > Wei Tang wrote: >> Hi, >> >> I found out there are some problems with octave-atlas 3.0.0-4, which >> can be compiled but will crash when using. Specifically, I tried the >> following commands: >> >> a = randn(700,700); >> b = a'*a; % <=== this line will cause octave crash. >> >> but if I only try a', which is ok. > > Indeed, I get this too. This looks like a crash inside some atlas > function, as the gdb output shows (the CrashReporter output is even > less > helpful): > > octave:2> b=a'*a; > > Program received signal EXC_BAD_ACCESS, Could not access memory. > Reason: 13 at address: 0x00000000 > [Switching to process 24424 thread 0x1003] > 0x028c1391 in ATL_dupMBmm0_6_0_b0 () > (gdb) bt > #0 0x028c1391 in ATL_dupMBmm0_6_0_b0 () > #1 0x057bda20 in ?? () > Cannot access memory at address 0x502 > > No idea how to debug this. >
No idea either ... 1) here the commands go through w/o problem (this is on a CoreDuo, I'll re-try tomorrow on a G5, both still running 10.4). To both of you, on what machine was this ? 2) grepping even just for the substring dupMBmm in atlas's files yields nothing : > # dpkg -L atlas atlas-doc atlas-shlibs|xargs file|fgrep -v 'directory > symbolic'|cut -f1 -d:|xargs fgrep dupMBmm > # and same thing in the build-dir ... JF Mertens ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ Fink-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/fink-users
