> On Nov 1, 2016, at 1:22 PM, Matthew Knepley <[email protected]> wrote:
> 
> On Tue, Nov 1, 2016 at 7:19 AM, Garth N. Wells <[email protected] 
> <mailto:[email protected]>> wrote:
> I get a PaStiX build failure with '--download-pastix':
> 
> In file included from sopalin/src/murge.c:364:0,
>                  from sopalin/src/zmurge.c:54:
> sopalin/src/variable_csc.c: In function ‘vcsc_add_node’:
> sopalin/src/variable_csc.c:138:13: error: non-floating-point argument
> in call to function ‘__builtin_isnan’
>              if (isnan(vcsc->values[COL-1][i*dof2 + ii])) {
>              ^~
> sopalin/src/variable_csc.c: In function ‘vcsc_add’:
> sopalin/src/variable_csc.c:255:11: error: non-floating-point argument
> in call to function ‘__builtin_isnan’
>            if (isnan(vcsc->values[COL_NODE][idx])) {
>            ^~
> sopalin/src/variable_csc.c: In function ‘vcsc_to_cscd’:
> sopalin/src/variable_csc.c:663:15: error: non-floating-point argument
> in call to function ‘__builtin_isnan’
>                if (!isnan(vcsc->values[MYCOL-1][dof2*iter2+iterdof])) {
>                ^~
> make: *** [sopalin/obj/i686_pc_linux/zmurge.o] Error 1
> *******************************************************************************
> 
> From the PaStIX changelog, it looks like it's fixed in PaStiX 5.2.3.
> But, despite bumping the version in PaStiX.py cleaning everything out,
> PETSc kept downloading 5.2.2.20. I have no idea where the build system
> is getting 5.2.2.20 from.
> 
> From this line
> 
>   
> https://bitbucket.org/petsc/petsc/src/a2b5a0438a710504fe0fdd39a793ddf656611e6a/config/BuildSystem/config/packages/PaStiX.py?at=master&fileviewer=file-view-default#PaStiX.py-6
>  
> <https://bitbucket.org/petsc/petsc/src/a2b5a0438a710504fe0fdd39a793ddf656611e6a/config/BuildSystem/config/packages/PaStiX.py?at=master&fileviewer=file-view-default#PaStiX.py-6>
> 
> You can change that to download the version you want. If it works, let us 
> know and I will make a patch.

It does work (I’ve had this problem for months and I switched to 5.2.3 last 
week).

Thanks,
Pierre

>   Thanks,
> 
>     Matt
>  
> Also, the doc at
> 
>     
> http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MATSOLVERPASTIX.html
>  
> <http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MATSOLVERPASTIX.html>
> 
> seems inconsistent with config/BuildSystem/config/packages/PaStiX.py -
> PaStiX doesn't depend on ParMETIS and METIS.
> 
> Garth
> 
> 
> 
> -- 
> What most experimenters take for granted before they begin their experiments 
> is infinitely more interesting than any results to which their experiments 
> lead.
> -- Norbert Wiener

Reply via email to