My recent commit r186623 caused buildbots for s390-linux and ppc64-linux to fail. I rolled back that commit, and I’m trying to fix to re-submit.
Is there any good way to debug the issue which does not occur on my local x86-64 machine? I don’t have S390 nor PPC64 machines.
If it's failing on S390 and PPC64 in particular, your code probably
accidentally depends on the host endianness. (Looking at the commit
in question seems to confirm that.)
Yes, it’s very likely. It’d be very convenient if we could send a patch to buildbots for testing without actually submitting.
Even better, someone ought to implement a virtual machine architecture
that purposely broke every possible non-portable assumption in
unexpected and unpredictable ways, as an acid test environment for all
Chromium has long enjoyed the concept of ‘try-bots’ in addition to build bots. Thus far, the LLVM community hasn’t needed them badly enough for anyone to go through the pain of setting them up. Once you think you have the fix, just recommit and watch the bots.