Closed zbeekman closed 7 years ago
No, the correct fix is to change BASEDIR into INSTALLDIR. Rationale: the INSTALLDIR variable was copied from another subdirectory in tests, and I forgot to doublecheck; on my system it worked anyway because it was picking up a previous installation. The idea is that when building, it points to the build tree, but when (some of) the test subdirs are copied onto INSTALLDIR/Samples, they then point to the installed modules, not to the build directory. I have updated the Makefile accordingly, and have adjusted the tag. Thanks for the heads up.
Hi Salvatore,
Whatever the correct fix is, your make check
is still broken. The issue is that the run directory logic is broken too: in a new build it doesn't exist.
Also, just FYI, best practice for yanked releases is to mark them yanked and bump the patch version, just so people aren't confused.
In other subdirs the ./runs subdir is there with input files, so it does not need to be created anew. Can you fix it and bump the patch number? I have to rush away, will be fully online tomorrow.
Thanks Salvatore
On Thu, Oct 26, 2017 at 1:40 PM, zbeekman notifications@github.com wrote:
@zbeekman commented on this pull request.
In test/serial/Makefile https://github.com/sfilippone/psblas3/pull/1#discussion_r147128595:
$(FLINK) $(LINKOPT) d_matgen.o psb_d_xyz_mat_mod.o psb_d_xyz_impl.o \
-o d_matgen $(PSBLAS_LIB) $(LDLIBS) /bin/cp -p $(CPUPDFLAG) d_matgen $(EXEDIR)
/bin/mv d_matgen $(EXEDIR)
check: all
- cd runs && echo 5 | ./d_matgen
- cd $(EXEDIR) && echo 5 | ./d_matgen
what's the point of the EXEDIR variable if you don't use it?
— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/sfilippone/psblas3/pull/1#pullrequestreview-72176515, or mute the thread https://github.com/notifications/unsubscribe-auth/APX-T3gcX97Qg5g9--kqDbyR5xgXnt8tks5swH3PgaJpZM4QG07H .
make check
fails for silly reasons. This fixes the makefile in tests/serial.