https://bugs.kde.org/show_bug.cgi?id=438640
--- Comment #17 from Mark Wielaard <[email protected]> --- (In reply to mcermak from comment #16) > Created attachment 193219 [details] > updated patch > > Thank you Mark for your kind review. One of those "typos" as you say, is a > fail to pass the second argument to the instruction. No wonder the test > results were wrong. I've dropped my original testcase and added a new one > derived from its amd64 counterpart. Now the test results look much better > and expected. Local testing gives good results. New test result is > included and overall no regression spotted. > > Buildbot started two test runs: > https://builder.sourceware.org/buildbot/#/changes/120353 (sans NEWS entry) > https://builder.sourceware.org/buildbot/#/changes/120354 (incl NEWS entry) > > Please review. This looks good. BTW. your tests were good too. You don't have to add them to the new crc32 tests but could if you wanted to test more/differently. One small issues with: diff --git a/none/tests/x86/crc32.vgtest b/none/tests/x86/crc32.vgtest new file mode 100644 index 000000000..8ebf4ffa3 --- /dev/null +++ b/none/tests/x86/crc32.vgtest @@ -0,0 +1,3 @@ +prog: crc32 +prereq: ../../../tests/x86_amd64_features amd64-sse42 +vgopts: -q That prereq using x86_amd64_features only works on an amd64 (x86_64), so when running x86 as secondary arch. tests/x86_amd64_features isn't build on x86. So for now change it to: prereq: test -x crc32 So the test is only run when the crc32 binary is build. This is technically also wrong, but we'll fix that later. (It mimics what the sse4-x86 does. Which is also not build on x86. So we should fix both.) -- You are receiving this mail because: You are watching all bug changes.
