[Mesa-dev] Mesa 17.1.2 release candidate

Jason Ekstrand jason at jlekstrand.net
Sat Jun 3 18:43:31 UTC 2017


On Sat, Jun 3, 2017 at 11:30 AM, Juan A. Suarez Romero <jasuarez at igalia.com>
wrote:

> On Sat, 2017-06-03 at 11:10 -0700, Jason Ekstrand wrote:
>
> On Sat, Jun 3, 2017 at 10:05 AM, Juan A. Suarez Romero <
> jasuarez at igalia.com> wrote:
>
> On Thu, 2017-06-01 at 16:35 -0700, Jason Ekstrand wrote:
>
> On Thu, Jun 1, 2017 at 4:20 PM, Juan A. Suarez Romero <jasuarez at igalia.com
> > wrote:
>
> Hello list,
>
> The candidate for the Mesa 17.1.2 is now available. Currently we have:
>  - 51 queued
>  - 3 nominated (outstanding)
>  - and 9 rejected patch(es)
>
>
> The current queue contains fixes in several places.
>
> The current queue consists of:
>
> Several fixes in the autotools' configure that improves handling of target
> platforms.
>
> Mesa receives a fix for a leaking in a surface.
>
> OMX has a couple of fixes when building in a X11-less setup.
>
> ANV driver receives several fixes, like adding some missing transitions
> between layouts, or improvements in setting memory types and heaps.
>
> There are a few fixes in i965, R100, R200, Radeonsi/GFX9, Radv and
> Freedreno drivers.
>
> Also, EGL/Android receives a fix for a segfault that replaces a previous
> one that was not correct.
>
> Vulkan's Wayland gets some fixes, including the use of proxy wrappers for
> swapchains, or per-display event queue usage.
>
> Likewise EGL's Wayland now use per-surface event queues, as well as other
> fixes like ensuring back buffer is obtained.
>
> Take a look at section "Mesa stable queue" for more information.
>
>
> Testing reports/general approval
> --------------------------------
>
> Any testing reports (or general approval of the state of the branch)
> will be greatly appreciated.
>
> The plan is to have 17.1.2 this Friday (3rd of June), around or
> shortly after 23:00 GMT.
>
> If you have any questions or suggestions - be that about the current
> patch queue or otherwise, please go ahead.
>
>
> Trivial merge conflicts
> -----------------------
>
> commit e064f7d826812598749ce27868d9ff865c137ef4
> Author: Jason Ekstrand <jason.ekstrand at intel.com>
>
>     anv: Set up memory types and heaps during physical device init
>
>     (cherry picked from commit c1f4343807d1040bd7b5440aa2f5fccf5f12842d)
>
> commit ce2b96dd8b5a6c4ea0c80c6c0f46652a0b591515
> Author: Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
>
>     radv: Reserve space for descriptor and push constant user SGPR setting.
>
>     (cherry picked from commit 18efb404cfb38f722a16df7539390cf9a4a71929)
>
>
> Cheers,
> J.A.
>
>
> Mesa stable queue
> -----------------
>
> Nominated (3)
> =============
> Chad Versace(1):
>       i965/dri: Fix bad GL error in intel_create_winsys_renderbuffer()
>
> Jason Ekstrand(1):
>       i965: Rework Sandy Bridge HiZ and stencil layouts
>
>
> This just landed in master.
>
>
>
>
> Ok. I've included it in the Queued.
>
>
> Nicolas Dechesne(1):
>       util/rand_xor: add missing include statements
>
>
> Queued (51)
> ===========
> Bartosz Tomczyk (1):
>       mesa: Avoid leaking surface in st_renderbuffer_delete
>
> Bas Nieuwenhuizen (1):
>       radv: Reserve space for descriptor and push constant user SGPR
> setting.
>
> Daniel Stone (7):
>       vulkan: Fix Wayland uninitialised registry
>       vulkan/wsi/wayland: Remove roundtrip when creating image
>       vulkan/wsi/wayland: Use per-display event queue
>       vulkan/wsi/wayland: Use proxy wrappers for swapchain
> Squashed with
>       vulkan/wsi/wayland: Fix proxy wrappers for swapchain recreation
>       egl/wayland: Don't open-code roundtrip
>       egl/wayland: Use per-surface event queues
> Squashed with
>       egl/wayland: verify event queue was allocated
>       egl/wayland: Ensure we get a back buffer
>
> Emil Velikov (24):
>       docs: add sha256 checksums for 17.1.1
>       configure: move platform handling further up
>       configure: rename remaining HAVE_EGL_PLATFORM_* guards
>       configure: update remaining --with-egl-platforms references
>       configure: loosen --with-platforms heuristics
>       configure: enable the surfaceless platform by default
>       configure: set HAVE_foo_PLATFORM as applicable
>       configure: error out when building GLX w/o the X11 platform
>       configure: check once for DRI3 dependencies
> Squashed with
>       configure.ac: add xcb-fixes to the XCB DRI3 list
> Squashed with
>       configure.ac: s/xcb-fixes/xcb-xfixes/
>       loader: build libloader_dri3_helper.la only with HAVE_PLATFORM_X11
>       configure: error out when building X11 Vulkan without DRI3
>       auxiliary/vl: use vl_*_screen_create stubs when building w/o platform
>       st/va: fix misplaced closing bracket
>       st/omx: remove unneeded X11 include
>       st/omx: fix building against X11-less setups
>       gallium/targets: link against XCB only as needed
>       configure: error out if building VA w/o supported platform
>       configure: error out if building OMX w/o supported platform
>       configure: error out if building VDPAU w/o supported platform
>       configure: error out if building XVMC w/o supported platform
>       travis: remove workarounds for the Vulkan target
>       anv: automake: list shared libraries after the static ones
>       radv: automake: list shared libraries after the static ones
>       egl/wayland: select the format based on the interface used
>
> Ian Romanick (3):
>       r100: Don't assume that the base mipmap of a texture exists
>       r100,r200: Don't assume glVisual is non-NULL during context creation
>       r100: Use _mesa_get_format_base_format in radeon_update_wrapper
>
> Jason Ekstrand (8):
>       anv: Handle color layout transitions from the UNINITIALIZED layout
>       anv: Handle transitioning depth from UNDEFINED to other layouts
>       anv/image: Get rid of the memset(aux, 0, sizeof(aux)) hack
>       anv: Predicate 48bit support on gen >= 8
>       anv: Set up memory types and heaps during physical device init
>       anv: Set image memory types based on the type count
>       i965/blorp: Do and end-of-pipe sync on both sides of fast-clear ops
>       i965: Round copy size to the nearest block in intel_miptree_copy
>
> Juan A. Suarez Romero (1):
>       Revert "android: fix segfault within swap_buffers"
>
> Marek Olšák (1):
>       radeonsi/gfx9: compile shaders with +xnack
>
> Nicolai Hähnle (1):
>       st/mesa: remove redundant stfb->iface checks
>
> Nicolas Boichat (1):
>       configure.ac: Also match -androideabi tuple
>
> Rob Clark (1):
>       freedreno: fix fence creation fail if no rendering
>
> Tapani Pälli (1):
>       egl/android: fix segfault within swap_buffers
>
> Timothy Arceri (1):
>       st/mesa: don't mark the program as in cache_fallback when there is
> cache miss
>
>
> Rejected (9)
> ============
> Dave Airlie(1):
>        radv: fix regression in descriptor set freeing.
>
> This commit addressed an earlier commit 126d5ad which did not land in
> branch.
>
> Jason Ekstrand(7):
>        anv: Determine the type of mapping based on type metadata
>        anv: Stop setting BO flags in bo_init_new
>        anv: Make supports_48bit_addresses a heap property
>        anv: Advertise both 32-bit and 48-bit heaps when we have enough
> memory
>        anv: Require vertex buffers to come from a 32-bit heap
>        anv: Add valid_bufer_usage to the memory type metadata
>        anv: Refactor memory type setup
>
> All these patches have dependencies among them, and also about commits
> which did not
> land in branch, like c3c61d210f or 35e626bd0e.
>
>
> c3c61d210f should be back-ported to stable as well.  I'm aware of the
> dependency on 35e626bd0e but it should be fairly easy to rebase on top of
> it by just dropping all of the EXEC_ASYNC bits from "anv: Stop setting BO
> flags in bo_init_new".  I'm happy to do the rebase for you if you'd like.
>
>
>
> Thanks for info, Jason.
>
> I've been working hard to try to include them in Queued list. Finally, I
> added all of the ANV commits, except the one entitled "anv: Require vertex
> buffers to come from a 32-bit heap". I keep this one in the Rejected list
> because it was causing a regression in the Crucible test suite. The other 6
> are included in queued.
>
>
> Uh... That's the important one.  All of the others lead up to it.  What is
> the crucible regression?  I was pretty sure crucible master (you may have
> to pull) was working fine.
>
>
>
> Mark reported around 50 crashes in the crucible testsuite.
>
> Some of them are:
>
> crucible.func.depthstencil.basic-depth.clear-0.0.op-greater
> crucible.func.miptree.d32-sfloat.aspect-depth.view-2d.
> levels01.array01.extent-1024x512.upload-copy-from-
> buffer.download-copy-with-draw
> crucible.func.push-constants.basic
> crucible.func.shader.fragcoord
>
>
>
> Now, you're right: using Crucible master indeed it is working fine (at
> least on my local host).
>

Yeah, crucible needed a lot of fixing to work with the change.  It had some
very hard-coded assumptions about memory allocation that were no longer
true once we started requiring VBs to come from the 32-bit heap.


> I guess Mark is using instead commit
> 4f60bf1,
> which indeed has those regressions. But couldn't confirm with Mark.
>
>
> In this case, I can include the patch in the queue, ask Mark to update the
> used crucible test, and re-try again. If everything goes fine, I'll do the
> release. As we are in weekend, I guess I won't get the results until Monday.
>
> Does this sound fine for you?
>

Sounds good.


>
> J.A.
>
>
>
>
> Additionally, I've enqueued the one entitled "anv: Set EXEC_OBJECT_ASYNC
> when available" squashed with "anv/tests: Create a dummy instance as well
> as device".
>
> All those new commits caused the following new trivial merge conflicts:
>
> commit 2562b3252b63f2e938d6e6614e1f9a2fa2064140
> Author: Jason Ekstrand <jason.ekstrand at intel.com>
>
>     anv: Make supports_48bit_addresses a heap property
>
>     (cherry picked from commit b83b1af6f6936f36db42a8f8b8e0854d0f9491fd)
>
> commit 86a8854b118fd42d157a5b640fa7a3a1cad92301
> Author: Jason Ekstrand <jason.ekstrand at intel.com>
>
>     anv: Stop setting BO flags in bo_init_new
>
>     (cherry picked from commit 00df1cd9d6234cdfc9fb2bf3615196ff83a3c956)
>
> commit 0f042901e3cd451f9a7630376083a805328aebe3
> Author: Jason Ekstrand <jason.ekstrand at intel.com>
>
>     anv: Add valid_bufer_usage to the memory type metadata
>
>     (cherry picked from commit f7736ccf53eaeb66c4270afe0916e2cb29ab8667)
>
> commit 15bc6d4d210eee051407a816811012eba0a3be3b
> Author: Jason Ekstrand <jason.ekstrand at intel.com>
>
>     anv: Determine the type of mapping based on type metadata
>
>     (cherry picked from commit 92325a7efc769c32e03031323e21700dc55171e4)
>
>
> All these news changes implies that I need to wait for Intel CI results to
> ensure nothing is broken, which means I need to delay the release
> probably until Monday (or sooner if I get the results before).
>
> Sorry for the inconveniences!
>
>
>
> >
>
> Nicolai Hähnle(1):
>        radeonsi: load patch_id for TES-as-ES when exporting for PS
>
> This commit depends on 9fd9a7d0ba3 and 678d568c7b2, neither of which is in
> branch.
>
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
>
> _______________________________________________
> mesa-dev mailing listmesa-dev at lists.freedesktop.orghttps://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
>
>
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20170603/3b960626/attachment-0001.html>


More information about the mesa-dev mailing list