<html> <head> <base href="https://bugs.freedesktop.org/" /> </head> <body> <div> <a class="bz_bug_link bz_status_REOPENED " title="REOPENED - Uninitialized variables leads to random segfaults (valgrind log, apitrace attached)" href="https://bugs.freedesktop.org/show_bug.cgi?id=94955#c12">Comment # 12</a> on <a class="bz_bug_link bz_status_REOPENED " title="REOPENED - Uninitialized variables leads to random segfaults (valgrind log, apitrace attached)" href="https://bugs.freedesktop.org/show_bug.cgi?id=94955">bug 94955</a> from <a class="email" href="mailto:sroland@vmware.com" title="Roland Scheidegger <sroland@vmware.com>"> Roland Scheidegger</a> <pre>(In reply to Bruce Cherniak from <a href="show_bug.cgi?id=94955#c10">comment #10</a>) > Does the gallium llvm sampler fully support float textures? Both llvmpipe > and OpenSWR generate the same sampler and fail identically. Yes, that's easily supported (it's actually one of the most easy formats code-wise). The sampler in question doesn't even use any filtering, so compared to some other generated sampler funcs it's dead simple. Forcing tex func is actually a really good idea to separate things out (even more so when looking at the assembly), albeit with a float type it should not have been necessary (the logic there what's complex or not is quite flawed actually...) Though it doesn't look like float format to me. More like D24X8. > > I'll attach the sampler ir, if it's useful. Any hints on debugging sampler > jit are appreciated. :-) Any evidence it actually crashed there? Kind of difficult to debug if you don't get a valid address... My guess would be it's sampling the system (so not fbo) depth buffer, and the buffer disappeared in the meantime (if it really is crashing there)... In that case, it might be similar to <a class="bz_bug_link bz_status_NEW " title="NEW - llvmpipe crash in rendering on Atom" href="show_bug.cgi?id=94522">bug 94522</a>.</pre> </div> <hr> You are receiving this mail because: <ul> <li>You are the assignee for the bug.</li> <li>You are the QA Contact for the bug.</li> </ul> </body> </html>