This is the mail archive of the mailing list for the glibc project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

elf/tst-protected1[ab] failures

These tests fail (at least on i686, x86_64, and powerpc64, that I know of;
probably on all machines) when using all released versions of binutils.

The state in the libc tree should never be to expect that unreleased tools
are the norm.  These tests ought to be either marked for XFAIL or disabled
entirely when not using a linker that makes them succeed.  We should never
have allowed this state to exist in a libc release (release managers take

Can you figure out if it's possible to improve the existing configure check
controlling have-protected-data so that it will accept only linkers that
actually support these tests correctly?  If that's not feasible in the
configure check, then we'll just need to mark the tests for XFAIL.  It
would be nice if that could be conditional, but conditionalizing it on the
ld version string seems nasty, so perhaps it's better to just have it as
always XFAIL unless an empirical configure check can rule it out.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]