On Thu, Jun 20, 2019 at 03:03:33PM -0600, Thomas Frohwein wrote: > On Thu, Jun 20, 2019 at 05:22:53PM +1000, Jonathan Gray wrote: > [...] > > > I hit an abort trap with coredump on both Skylake and Vega 64 (the > > > latter w/ amdgpu). I don't recall an issue with vulkaninfo with sdk > > > version 1.1.104. vkcube still runs fine on both. Will take a deeper dive > > > to see what's going on with vulkaninfo that I missed... > > > > I see the same problem described here on broadwell, vkcube also runs. > > Linux seems to be more forgiving of getting pthreads wrong. > > > > Perhaps we revert > > https://github.com/KhronosGroup/Vulkan-Loader/commit/ecb0b1e69fb2f4d3cae262e6da24c170ce62ae13 > > ? > [...] > > Thanks for the hint; that fixed vulkaninfo on both my Skylake and Vega! > Now getting full vulkaninfo on both without any segfault or abort trap. > Also tested vkcupe and vkcubepp, as well as vkquake on the Vega + > amdgpu... and emulators/dolphin with vulkan backend on the Skylake :] > > Updated tarball attached that reverts the above commit and has the > missing MAINTAINER line for vulkan-headers. > > ok?
ok jsg@ I still get a crash on vkcube with ivy bridge but I think this is worth getting in. We should probably open a github issue for the loader pthreads problem.
