[Buildroot] libdrm: buildroot patchstack - static linking

[Buildroot] libdrm: buildroot patchstack - static linking

[Peter Seiderer]

Hello *,

some progress on up-streaming the buildroot libdrm patch stack:

 - 0001-Add-ARM-support-into-xf86drm.h.patch
   https://gitlab.freedesktop.org/mesa/drm/merge_requests/31
   Acked, but with one further question 'why not catch-all __ARM_ARCH?'
   Thomas - you touched the patch and added the comment, any further input?

 - 0002-xf86atomic-require-CAS-support-in-libatomic_ops.patch
   https://gitlab.freedesktop.org/mesa/drm/merge_requests/32
   Acked

 - 0003-meson.build-fix-intel-atomics-detection.patch
   merged https://gitlab.freedesktop.org/mesa/drm/merge_requests/27

 - 0004-meson.build-enable-static-build.patch
   declined https://gitlab.freedesktop.org/mesa/drm/merge_requests/28
   'Closing the MR, but feel free to reopen if it turns out someone has
   a real need for this'

   Anyone out there using libdrm static linking?

 - 0005-tests-meson.build-disable-nouveau-tests-for-static-b.patch
   folded into previous one https://gitlab.freedesktop.org/mesa/drm/merge_requests/27
   and declined

 - 0006-tests-nouveau-threaded-adapt-ioctl-signature.patch:
   merged https://gitlab.freedesktop.org/mesa/drm/merge_requests/30

Regards,
Peter

[Buildroot] libdrm: buildroot patchstack - static linking

[Thomas Petazzoni]

On Thu, 5 Dec 2019 21:54:53 +0100
Peter Seiderer <ps.report@gmx.net> wrote:

> some progress on up-streaming the buildroot libdrm patch stack:
> 
>  - 0001-Add-ARM-support-into-xf86drm.h.patch
>    https://gitlab.freedesktop.org/mesa/drm/merge_requests/31
>    Acked, but with one further question 'why not catch-all __ARM_ARCH?'
>    Thomas - you touched the patch and added the comment, any further input?

Well, my comment explains it all: ldrex/strex only exist on ARM >= v6,
but __ARM_ARCH is true for all ARM ISAs, including ARMv4 and ARMv5,
which Buildroot supports, but don't provide ldrex/strex.

>  - 0004-meson.build-enable-static-build.patch
>    declined https://gitlab.freedesktop.org/mesa/drm/merge_requests/28
>    'Closing the MR, but feel free to reopen if it turns out someone has
>    a real need for this'
> 
>    Anyone out there using libdrm static linking?

But what is the cost for upstream to use the correct library() Meson
macro/function instead of shared_library() ? It is shorter, and does
the job. I'm not sure why upstream is reluctant to merge this.

>  - 0005-tests-meson.build-disable-nouveau-tests-for-static-b.patch
>    folded into previous one https://gitlab.freedesktop.org/mesa/drm/merge_requests/27
>    and declined

So, they basically say they don't want to support static linking at all?

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

[Buildroot] libdrm: buildroot patchstack - static linking

[Peter Seiderer]

Hello Thomas,

On Thu, 5 Dec 2019 22:11:51 +0100, Thomas Petazzoni <thomas.petazzoni@bootlin.com> wrote:

> On Thu, 5 Dec 2019 21:54:53 +0100
> Peter Seiderer <ps.report@gmx.net> wrote:
>
> > some progress on up-streaming the buildroot libdrm patch stack:
> >
> >  - 0001-Add-ARM-support-into-xf86drm.h.patch
> >    https://gitlab.freedesktop.org/mesa/drm/merge_requests/31
> >    Acked, but with one further question 'why not catch-all __ARM_ARCH?'
> >    Thomas - you touched the patch and added the comment, any further input?
>
> Well, my comment explains it all: ldrex/strex only exist on ARM >= v6,
> but __ARM_ARCH is true for all ARM ISAs, including ARMv4 and ARMv5,
> which Buildroot supports, but don't provide ldrex/strex.

Thanks for backing-up my (already done) reply to the merge request (see link above),
will add an additional comment to the change set...

>
> >  - 0004-meson.build-enable-static-build.patch
> >    declined https://gitlab.freedesktop.org/mesa/drm/merge_requests/28
> >    'Closing the MR, but feel free to reopen if it turns out someone has
> >    a real need for this'
> >
> >    Anyone out there using libdrm static linking?
>
> But what is the cost for upstream to use the correct library() Meson
> macro/function instead of shared_library() ? It is shorter, and does
> the job. I'm not sure why upstream is reluctant to merge this.

Not sure about this change of mind (from Reviewed by to declined), the
original comment:

	On a technical level, this MR is fine and we can merge it, but the thing
	is that it adds to the maintenance burden (not much, sure) and limits
	future features because they will need to work with static linking,
	not to mention the usual problems that made everyone stop statically
	linking (although one could argue that this last point would be our
	users' choice, but should we give them a foot-gun and the choice to use it?).

	Unless there's a need by anyone to support this, I think we shouldn't do
	this "just because we can" :)

>
> >  - 0005-tests-meson.build-disable-nouveau-tests-for-static-b.patch
> >    folded into previous one https://gitlab.freedesktop.org/mesa/drm/merge_requests/27
> >    and declined
>
> So, they basically say they don't want to support static linking at all?

	Closing the MR, but feel free to reopen if it turns out someone has a real need for this

But open minded in case of a 'real world' example of usefulness of libdrm static linking...

Regards,
Peter

>
> Thomas