Bug#1050076: gtk4: 4.12 regression: FTBFS on i386: assertion failed (r == tests[i].r (+/- FLT_EPSILON))

[Simon McVittie]

Source: gtk4
Version: 4.12.0+ds-1
Severity: serious
Tags: ftbfs
Justification: fails to build from source (but built successfully in the past)
Usertags: i387

gtk4 4.12.0 in experimental has test failures on multiple buildds.
On i386 there is this failure:

not ok /color/roundtrips - ERROR:../../../testsuite/gtk/colorutils.c:43:test_roundtrips: assertion failed (r == tests[i].r (+/- FLT_EPSILON)): (1.1920929e-07 == 0)

I think this is our old enemy, the legacy i387 FPU and its 80-bit
extended precision, and the test will need to apply more fuzz to its
acceptable result.

If our i386 port had a baseline that included SSE2 so that we could
build everything with -mfpmath=sse, this class of bug would probably
go away; but at the moment our baseline is officially a not-quite-i686
(i686 with one missing opcode, to accommodate certain Geode CPUs) so
we cannot assume MMX, SSE or SSE2, and various developers have resisted
suggestions to raise the baseline. I'm going to start tracking bugs in
this class with a usertag to get an idea of how many of them there are.

    smcv