Skip Menu |

This queue is for tickets about the PDL-LinearAlgebra CPAN distribution.

Report information
The Basics
Id: 70572
Status: resolved
Priority: 0/
Queue: PDL-LinearAlgebra

People
Owner: chm [...] cpan.org
Requestors: sisyphus [...] cpan.org
Cc:
AdminCc:

Bug Information
Severity: Critical
Broken in: 0.06
Fixed in: 0.08



Subject: [Win32] Crash while running test suite
The crash occurs during the running of test 7 in t/1.t. Specifically, it occurs with the following zaxpy_() call in cgeexp(): zaxpy_(&mm, &z__1, &wsp[ip], &c__1, &wsp[iq], &c__1); I don't yet know why it happens, and with my lack of skill regarding use of a debugger it might take me a while. Any ideas ? I have a number of perls - in all instances the compiler I use is the compiler that built both perl and lapack (GotoBlas2). The 64-bit build of perl-5.14.0 (using gcc-4.7.0) is fine - I haven't yet tested the 64-bit builds on any other perl versions. But the 32-bit build inevitably crashers on both perl-5.12.0 and 5.14.0 (which are the only versions I've tried). With perl-5.12.0 I'm using gcc-3.4.5, with perl-5.14.0 I've tried both gcc-3.4.5 and gcc-4.5.2. Cheers, Rob
RT-Send-CC: vanuxemg [...] yahoo.fr
On Mon Aug 29 04:35:29 2011, SISYPHUS wrote: Show quoted text
> The crash occurs during the running of test 7 in t/1.t. > Specifically, it occurs with the following zaxpy_() call in cgeexp(): > > zaxpy_(&mm, &z__1, &wsp[ip], &c__1, &wsp[iq], &c__1);
I've just compared the values of mm, z__1, wsp[ip], c__1 and wsp[iq] between the 32-bit build and the 64-bit build. Afaict they are identical - which leaves me thinking that it's either a bug in the 32-bit build of GotoBLAS2, or a bug in the 3.4.5 and 4.5.2 mingw.org compilers. Cheers, Rob
Rob-

Is this ticket still valid?  I'm trying to close outstanding PDL::LInearAlgebra issues.

Thanks,
Chris
On Wed Dec 04 09:58:47 2013, CHM wrote: Show quoted text
> Is this ticket still valid?
I think it was bugginess in the blas library I was using. I've since switched to a different blas implementation, at which point the problem went away. I've no objection of you want to close this ticket. Cheers, Rob