fix: ppc64el test failures in tst_disassemblyoutput#768
Conversation
Two fixes for ppc64el (powerpc64le) test failures: 1. readelf -s output format differs on ppc64el — the regex character class [0-9 a-zA-Z]+ is too restrictive. Changed to [^\n]+ which matches any format while still stopping at the newline. 2. Not every architecture emits branch visualisation for every disassembly line. On ppc64el some lines have empty visualisation. Wrapped the QVERIFY with an isEmpty() check so the character validation only runs when visualisation is present. Both fixes from the Debian PowerPC porters analysis in https://bugs.debian.org/1129621#17 Fixes KDAB#766
|
Hey there, sorry for the delay @Scottcjn. Can you get us an objdump text output that we can use for extending our test coverage? See |
|
@milianw — generated objdump output for ppc64el (POWER8 hardware: IBM POWER S824, Ubuntu 20.04, GCC 10.5). Binary: Verified header: First 300 lines of If you need different sections (full For
|
Fixes two test failures on ppc64el in tst_disassemblyoutput:
readelf regex: Character class [0-9 a-zA-Z]+ too restrictive for ppc64el readelf -s output. Changed to [^\n]+ which matches any format while stopping at newline.
Branch visualisation: Not every architecture emits branch visualisation for every disassembly line. Wrapped QVERIFY with isEmpty() check.
Both fixes from the Debian PowerPC porter analysis in https://bugs.debian.org/1129621#17
Tested on IBM POWER8 S824 (ppc64le). The regex change is backwards-compatible.
Fixes #766