Open tim77 opened 3 years ago
That could only happen at https://github.com/ValvePython/vdf/blob/master/vdf/__init__.py#L361
Which means that utf-16
on that build host must have been understood as utf-16be
(platform dependent), and since the test uses LE there will be a mismatch. Looking at https://kojipkgs.fedoraproject.org//work/tasks/713/60470713/hw_info.log confirms it.
I'd say this is a bug and I suppose the temporary fix is set the decode to utf-16le
. I'm yet to see a single example of Valve using this type anywhere. It's all utf-8
. Now that I think of it, I suppose all encode/decode should be LE, otherwise it all breaks on a BE system.
utf-16 on that build host must have been understood as utf-16be (platform dependent)
@tim77, if you're reporting a build/test failure on an unusual CPU architecture, please say so up-front? Otherwise maintainers will tend to assume that you were building on x86.
Debian s390x (which is big-endian) is exhibiting the same build failure (after I fixed a packaging bug that meant we weren't actually running the tests as we intended to). I'll send a merge request shortly.
Some tests not passed in Fedora 32:
But tests passed in Fedora 33 and no such issue. Full F32 build log.
Additional information:
Fedora 32
3.8.7