Mesa 19.1.0 RC 5 and stable branch 19.0.6 available, 19.1.0 final is delayed
The who blocker bugs which has been in Mesa 19.1.0 release candidates since RC2 remain open. The latest stable branch release 19.0.6 was meant to be the last of the 19.0.x series but 19.1.0 is delayed due to the release-blocking bugs so there will be one last stable 19.0.7-release before the final version of 19.1.0 becomes available.
See our earlier article Mesa 19.1.0 rc1 released with many new OpenGL features for an overview of what will be new in Mesa 19.1.0. The release-candidates since then have mainly been bug-fixes.
The stable branch release 19.0.6 fixes the following bugs in the 19.0.x series:
- gallium/swr: Param defaults for unhandled PIPE_CAPs
- gallium/aux: add PIPE_CAP_MAX_VARYINGS to u_screen
- nir: Actually propagate progress in nir_opt_move_load_ubo.
- anv/android: fix missing dependencies issue during parallel build
- etnaviv: use the correct uniform dirty bits
- Revert "mesa: unreference current winsys buffers when unbinding winsys buffers"
- winsys/drm: Fix out of scope variable usage
- docs: Add Sha256 sums for 19.0.5
- cherry-ignore: Add a commit that was manually backported
- cherry-ignore: add another 19.1 only patch
- cherry-ignore: add another 19.1 only patch
- gallium: wrap u_screen in extern "C" for c++
- VERSION: bump to 19.0.6
- docs: Add relnotes for 19.0.6
- egl/dri: flesh out and use dri2_create_drawable()
- swr/rast: fix 32-bit compilation on Linux
- vulkan: fix build dependency issue with generated files
- u_blitter: don't fail mipmap generation for depth formats containing stencil
- ac: fix a typo in ac_build_wg_scan_bottom
- etnaviv: fill missing offset in etna_resource_get_handle
- freedreno/ir3: dynamic UBO indexing vs 64b pointers
- freedreno/ir3: set more barrier bits
- freedreno/a6xx: fix GPU crash on small render targets
- intel/compiler: Fix assertions in brw_alu3
- radv: allocate more space in the CS when emitting events
- radv: do not use gfx fast depth clears for layered depth/stencil images
- Revert "st/mesa: expose 0 shader binary formats for compat profiles for Qt"
- st/glsl: make sure to propagate initialisers to driver storage
The latest RC5 release-candidate in the upcoming Mesa 19.1.0 did NOT close these release-blocking bugs:
- #110302 - [bisected][regression] piglit egl-create-pbuffer-surface and egl-gl-colorspace regressions
- #110357 - [REGRESSION] [BISECTED] [OpenGL CTS] cts-runner --type=gl46 fails in new attempted "41" configuration
However, there were bugs fixes and improvements made. These were:
- d3dadapter9: Revert to old throttling limit value
- nir: Actually propagate progress in nir_opt_move_load_ubo.
- swr/rast: fix 32-bit compilation on Linux
- iris: Don't assume UBO indices are constant
- intel/fs,vec4: Use g0 as the header for MFENCE
- intel/fs: Do a stalling MFENCE in endInvocationInterlock()
- nir/dead_cf: Call instructions aren't dead
- freedreno/ir3: fix input ncomp for vertex shaders
- Update version to 19.1.0-rc5
- nir/lower_non_uniform: safely iterate over blocks
- u_blitter: don't fail mipmap generation for depth formats containing stencil
- ac: fix a typo in ac_build_wg_scan_bottom
- radeonsi: init sctx->dma_copy before using it
- ac/nir: mark some texture intrinsics as convergent
- freedreno/ir3: set more barrier bits
- freedreno/a6xx: fix GPU crash on small render targets
- intel/compiler: Fix assertions in brw_alu3
- radv: allocate more space in the CS when emitting events
- radv: do not use gfx fast depth clears for layered depth/stencil images
- st/glsl: make sure to propagate initialisers to driver storage
- freedreno: Fix GCC build error.
Nothing really stands out, there's just the usual incremental improvements. The blocker-bugs which have remained open for almost a month remain unsolved. They are being worked on and there is a fair chance Mesa 19.1.0 won't be delayed for that much longer. It is clear that it will not be released on schedule.
published 2019-06-06 - last edited 2019-06-26
Enable comment auto-refresher