Marti Raudsepp <marti@juffo.org> writes:
> We're already seeing first buildfarm failures, on system "narwhal"
> using an msys/mingw compiler.
Yeah. After a full day's cycle, the answer seems to be that
denormalized input works fine everywhere except:
1. mingw on Windows (even though MSVC builds work)
2. some Gentoo builds fail (knowing that platform, the phase of the moon
is probably the most predictable factor determining this).
I'm inclined at this point to remove the regression test cases, because
we have no principled way to deal with case #2. We could install
alternative "expected" files that accept the failure, but then what's
the point of having the test case at all?
It might be worth pressuring the mingw folk to get with the program,
but I'm not going to be the one to do that.
regards, tom lane