This request is superseded by
request 1217399
(Show diff)
Overview
Request 1207485 superseded
- Update to version 7.0.0
* To avoid name clashes, use type "singlecomplex" instead of
"complex".
* Rank Deficient Matrices
- Created by mathletic
- In state superseded
- Superseded by 1217399
Loading...
Request History
mathletic created request
- Update to version 7.0.0
* To avoid name clashes, use type "singlecomplex" instead of
"complex".
* Rank Deficient Matrices
Wonder if this is what causes the hpc builds to fail because we have for the corresponding openblas flavor,
%undefine c_f_ver
still.I looked up the spec file of openblas and it is done there similar. What would you propose to do instead?
Sorry, my mistake, I do see that they have similar macros defined.
Could you try adding
%hpc_libdir
to the linker path before running the tests?@eeich, any better ideas?
I added export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:%hpc_libdir" right before the make command. It does not work. Is this what you asked me to do? Code is in home:mathletic:branches:science / superlu
No, it fails at
%check
. Adding it at themake
stage does not help with goes on later at%check
. Try adding it just before running%ctest
.Remained not working :-(
Sorry for the delay, but I sent you a fix (sr#1217347) just now. Hope that helps.
I don't know why the gnu-hpc packages fail. Neither I nor SuperLU did any change regarding openblas.
If you
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/lib/hpc/gnu14/openblas/0.3.27/lib64/
before %ctest, tests won't fail. The real challenge is making this less fragile.@Dmitry_R, @StefanBruens, @TheBlackCat, @adrianSuSE, @anag, @badshah400, @dstoecker, @eeich, @kwk, @lrupp, @mslacken, @openfoam, @psmt: review reminder