FTBFS of package widelands, possibly linked to the handling of UTF?

[Martin Quinson <Martin.Quinson@ens-rennes.fr>]

Hello, (please CC me in answers, as I'm not on the list)

The latest version of Widelands breaks on s390x:
https://buildd.debian.org/status/fetch.php?pkg=widelands&arch=s390x&ver=2%3A1.2-1&stamp=1717286193&raw=0

I reported the issue upstream, and they are pointing me to the following:
https://github.com/widelands/widelands/issues/6468

<<<<<<<<<<<<<<<<<<<<
The many format logs are correct, these come from testcases that check that
invalid format sequences are handled in a well-defined way. The real error is
in the checksum test:

TESTCASE(checksum) {
	const char* const text = "Hello World! This is a string with Ûñīcøđȩ
Bÿtèş.";
	SimpleMD5Checksum md5sum;
	md5sum.data(text, strlen(text));
	md5sum.finish_checksum();

	check_equal(md5sum.get_checksum().str(),
"d4e32b0d5b4fc7b10c7c46fafabf1e17");
}

I don't know the specifics of s390, so maybe one of these points might apply…:

    Is the string literal encoded as UTF-8 by default? (seems the most likely
problem to me)
    Are byte and fixed-width integer sizes guaranteed to be "normal" (1 byte =
8 bits, int32_t = 32 bits = 4 bytes, etc)? Also, what's the endianness of this
architecture?
>>>>>>>>>>>>>>>>>>>>>>

I'm very unfamiliar with s390x, and would appreciate if someone could help me
answering these questions on the github ticket or otherwise, so that we can
help upstream fixing the issue, please.

Thanks in advance,
Mt

Attachment: signature.asc
Description: This is a digitally signed message part