Package Details: libdrm-git 2.4.110.r27.g83763622-1

Git Clone URL: (read-only, click to copy)
Package Base: libdrm-git
Description: Userspace interface to kernel DRM services, master git version
Upstream URL:
Licenses: custom
Conflicts: libdrm
Provides: libdrm
Submitter: Lone_Wolf
Maintainer: Lone_Wolf
Last Packager: Lone_Wolf
Votes: 9
Popularity: 0.000176
First Submitted: 2015-12-23 15:03 (UTC)
Last Updated: 2022-02-23 22:17 (UTC)

Required by (244)

Sources (2)

Latest Comments

Lone_Wolf commented on 2021-08-06 10:26 (UTC)

source URI adjusted.

Using an existing git source tree will fail, Either change origin manually OR remove the old tree.

acidunit commented on 2021-08-04 19:12 (UTC)

The default branch has been renamed to "main" instead of "master".

shoober420 commented on 2020-09-20 19:29 (UTC)

Thank you for this package.

Lone_Wolf commented on 2019-12-20 09:37 (UTC)

Thanks for the details.

I will test adding that soon , probably this weekend, and upload a new version if there are no downsides.

waschtl commented on 2019-12-20 00:40 (UTC)

I forgot to mention the simple change needed in PKGBUILD for the "simple fix" (workaround) I mentioned:

check() {
   # '-t 10' is needed for the 'threaded' test, which uses the default meson
   # test timeout of 30 seconds. This is too short for many systems. It can be
   # removed if upstream fixes the issue.
   meson test -C _build -t 10

I agree, the real issue lies upstream. This workaround helped me get the package compiled, though, so I thought I would share my experience here. I agree with your suspicion about the test only running with an nvidia card in the system (my system has one).

Lone_Wolf commented on 2019-12-19 21:37 (UTC)

On my system that test is skipped very fast.

from libdrm-git/src/_build/meson-logs/testlog.txt

12/16 threaded                                SKIP     0.01 s 

--- command ---
--- stderr ---
Opening nouveau render node failed with -1

Perhaps that test only runs on systems with nvidia cards ? I do think this is an upstream issue, maybe report it on the dri-devel mailing list.

waschtl commented on 2019-12-18 21:58 (UTC)

On my system one of the checks (threaded) took longer than meson's default of 30s. This is a problem that was also noticed by the LFS guys [1].

On my system, the test completed successfully after 68.48s. It is a fairly old system (Intel Core i3-380M from 2010), but probably not the oldest to be used when compiling this package.

The simple fix would be to add a timeout multiplier to the meson test command. A multiplier of 10 will increase the timeout to 300s, which should be enough for far older systems, as well.

The more complex fix would be to patch libdrm's threaded test so that it has its own timeout value, like the random test probably does (as it runs longer with no issues).


Lone_Wolf commented on 2019-05-23 10:23 (UTC) (edited on 2019-05-23 10:24 (UTC) by Lone_Wolf)

When I started this package I used the licensing stuff from extra/libdrm.

I've looked at the libdrm source tree and there's no general licensing statement. Almost every invidual file has it's own copyright section in them.

Even if all files fall under the MIT license now, we'd have to extract the copyright notices with the names of the copyright holders and put that in one or more files in the /usr/share/licenses/$pkgname .

We know now that atleast libdrm, lib32-libdrm , libdrm-git , lib32-libdrm-git and amdgpu-pro have to deal with this, there may be others.

I think this should be discussed on the arch-general mailing list. Please start a topic there about this, Ashark.

Ashark commented on 2019-05-22 17:37 (UTC)

Hi. Shouldn't this package's license be changed to MIT? And also for libdrm that's in Extra. I am asking because I am repacking amdgpu-pro and there is also such situation (several MIT texts).

Lone_Wolf commented on 2018-09-02 11:51 (UTC)

Upstream solved the test failure they introduced, no need to use makepkg --nocheck anymore.

SR_team commented on 2018-08-12 14:38 (UTC)

11/16 amdgpu-symbol-check FAIL 0.42 s (exit status 1)

Lone_Wolf commented on 2017-08-14 11:36 (UTC)

Without bcunit-cunit-compat installed several (mostly admgpu related) tests are skipped during check() . No idea why stock libdrm package doesn't perform those tests.

pschichtel commented on 2017-08-13 18:02 (UTC)

why does this package depend on bcunit-cunit-compat even though the main libdrm package does not?

Lone_Wolf commented on 2017-04-05 22:02 (UTC) (edited on 2017-04-05 22:02 (UTC) by Lone_Wolf)

Thanks for the alert, damien. PKGBUILD updated

damien commented on 2017-04-05 11:22 (UTC)

The sed line to remove pthread-stubs breaks the configure script now. It should no longer be needed since a recent commit (63120172668610cebfd01e89f67e73037037dfd9) changes the pthread-stubs requirement to only apply on BSD platforms.

Lone_Wolf commented on 2016-11-13 16:50 (UTC)

The version in provides was used to allow mesa-git to keep depending on libdrm when it needed the latest not-yet-released-version to support AMD R4xx cards. I'll prob remove the version in next PKGBUILD.

sandy8925 commented on 2016-11-10 17:12 (UTC)

Shouldn't this be updated to say "Provides: libdrm=2.4.71" ?