On Sat, Nov 4, 2017 at 7:20 PM, Bruno Haible <[email protected]> wrote: > Hi Paul, > >> This is due to a bug in the Linux kernel, when it emulates 32-bit Linux atop >> a >> 64-bit kernel. See: >> >> https://bugzilla.redhat.com/show_bug.cgi?id=1419736 > > I see, you nailed it down already. > >> I see that Assaf reported the same bug against gzip here: >> >> https://bugs.gnu.org/25636#8 >> >> I don't see an easy way of working around the bug in gzip proper. It will >> become >> a bigger deal when the year 2038 rolls around.... >> >> I suppose we could skip the test when running in 32-bit mode atop Linux >> x86-64, >> though that sort of misses the point of doing the test. > > The concept of "expected failure" fits here, I think. If the test suite would > flag > this as an XFAIL rather than a FAIL, people would not report it. > > AFAIU from the doc [1], the way to turn a FAIL into an XFAIL is to add > something > like this in the Makefile.am: > > XFAIL_TESTS = > if $host is x86_64-*-linux* and $HOST_CPU_C_ABI is i386 > XFAIL_TESTS += timestamp > fi > > Bruno > > [1] > https://www.gnu.org/software/automake/manual/html_node/Scripts_002dbased-Testsuites.html
Thanks to both of you. FYI, the timestamp test also fails on e.g., the Solaris 5.11 i386 platform.
