Closed oodler577 closed 3 years ago
This is a great start! A few points:
CONFIGURE_REQUIRES
libomp-dev
if gcc
bit (which obviously can be expanded for other possibilities) needs an else
that does os_unsupported
check_lib
bit will need a function
as previously discussed to ensure actual OpenMP per the standard (that could go in the else
as a Hail Mary) - this implies starting with a %check_lib_args
of defaults which get added to in the if
branches, then check_lib
gets run once with it (achieving DRY)is_deeply
on the interpretationsTotals | |
---|---|
Change from base Build 757475001: | 100.0% |
Covered Lines: | 19 |
Relevant Lines: | 19 |
Note on one of your comments above, @mohawk2 -
the test shows a code-driven approach, and a more sustainable one is to separate the environment-info-gathering phase, from the info-interpretation phase; with that, a test could run anywhere since it would just have a hardcoded list of infos, and do is_deeply on the interpretations
If you are suggesting maintaining a set of profiles, I'd prefer to way until we might actually need it. The current approach simply updates a localized %Config::Config
by inserting compiler_name
as desired. I am open to expanding this in either case. The test is meant to be decoupled from the environment. If you can show me how it's not, I'd be happy make sure that it's fully independent as such. Since I do agree with your main point, which I read as, "the tests should run everywhere regardless of the environment."
still working down the list:)
@mohawk2 I expect the changes to pass. I believe I have addressed all issues you presented. Please let me know if there is anything else. I'd like to add you under 'AUTHORS', please tell me how you'd like that listed as one final change.
@mohawk2 - no rush, but I'd like to get these changes finalized today. I am eager to begin using this modules for other things. I also have quite a few "Alien" modules I think I might be able to knock out for (HPC related mostly). Anyway, just wanted to finish so I can recover any bandwidth and even utlize this for the talk I am preparing for the end of the month. Thanks for your continued guidence!
There is still no separate info-gather and info-evaluating phases which would enable the "basic" test to be data-driven. I will merge and release this anyway.
cflags -> -fopenmp ldflags -> -fopenmp
Makefile.PL also uses Devel::CheckLib to verify we're in a sane gcc environment when ccname == 'gcc'.