Removing intel-specific sanity check for libdrm #9694
Conversation
|
On POWER9: |
|
Test report by @bear-rsg |
|
Test report by @bear-rsg |
|
Test report by @Flamefire |
|
@Flamefire Making the sanity check arch-specific is easy through a custom easyblock, may be worth doing that here? |
|
Test report by @boegel |
|
I was thinking about supporting that in the framework. We have arch specific EasyConfig settings already (checksum? filename? can't remember) so I'd rather allow |
That won't work in that exact way though, because we already have support for tuples in |
|
@Flamefire @edmondac Is there an equivalent to sanity_check_paths = {
'files': ['include/xf86drm.h', 'include/xf86drmMode.h',
('lib/libdrm_intel.%s' % SHLIB_EXT, 'lib/libdrm_power.%s' % SHLIB_EXT), |
|
Seemingly there is no equivalent. Tested using libdrm 2.4.99 in 2019b. On CascadeLake: $ ls -1 ${EBROOTLIBDRM}/lib/
libdrm_amdgpu.la
libdrm_amdgpu.so
libdrm_amdgpu.so.1
libdrm_amdgpu.so.1.0.0
libdrm_intel.la
libdrm_intel.so
libdrm_intel.so.1
libdrm_intel.so.1.0.0
libdrm.la
libdrm_nouveau.la
libdrm_nouveau.so
libdrm_nouveau.so.2
libdrm_nouveau.so.2.0.0
libdrm_radeon.la
libdrm_radeon.so
libdrm_radeon.so.1
libdrm_radeon.so.1.0.1
libdrm.so
libdrm.so.2
libdrm.so.2.4.0
libkms.la
libkms.so
libkms.so.1
libkms.so.1.0.0On POWER: $ ls -1 ${EBROOTLIBDRM}/lib/
libdrm_amdgpu.la
libdrm_amdgpu.so
libdrm_amdgpu.so.1
libdrm_amdgpu.so.1.0.0
libdrm.la
libdrm_nouveau.la
libdrm_nouveau.so
libdrm_nouveau.so.2
libdrm_nouveau.so.2.0.0
libdrm_radeon.la
libdrm_radeon.so
libdrm_radeon.so.1
libdrm_radeon.so.1.0.1
libdrm.so
libdrm.so.2
libdrm.so.2.4.0
libkms.la
libkms.so
libkms.so.1
libkms.so.1.0.0 |
|
Well, we could use |
this is always there. So if you list this as an alternative you could as well not check for the other alternative |
|
Test report by @boegel |
|
Test report by @boegel |
|
OK, so we have a couple of options here:
sanity_check_paths = {
'files': [{'arch=x86_64': 'lib/libdrm_intel.%s' % SHLIB)EXT}, ...],
'dirs': [...],
}That shouldn't be extremely difficult either I think, and it may be worthwhile in the long run, but it's a bit more work I guess. @Flamefire, @edmondac , @branfosj: pick your poison! :) |
|
My vote would be for 3). I guess one can use the logic implemented for |
Partially maybe, yes. |
|
I was going to try option 2 @boegel and @Flamefire - but I don't have time to do option 3. @Flamefire are you up for doing option 3 yourself? |
|
I started working on it but gave up as it got incredibly complicate after encountering the type checking of EC options. The type of the values for the I'll keep my work here: https://github.com/Flamefire/easybuild-framework/tree/alternative_sanity_check_paths but for now I suggest an EasyBlock |
|
OK - I'll work on option 2 :-) |
|
Changed libdrm to use the custom easyblock in easybuilders/easybuild-easyblocks#1983 |
|
Test report by @Flamefire |
|
I've got test reports building now, but they're making GCC 8.2.0 currently... |
|
Test report by @Flamefire |
|
Test report by @bear-rsg |
|
Test report by @bear-rsg |
|
Test report by @boegel |
|
Test report by @boegel |
|
Test report by @boegel |
|
re-triggered tests now easybuilders/easybuild-easyblocks#1983 is merged... |
…asyconfigs into 20200116163226_new_pr_libdrm2499
|
Test report by @boegel |
|
Going in, thanks @edmondac! |
(created using
eb --new-pr)This now requires easybuilders/easybuild-easyblocks#1983