On Sat, 12 Jul 2014 14:20:22 +0200, Bjorn Reese <bre...@mail1.stofanet.dk> wrote: [...] > Applied.
Thanks! > > It fixes it on all three platforms I'm looking at here: PowerPC, SPARC and > > S390x. > > Great. I have created a fix that marks PowerPC, SPARC, and S390x as > using double-double, and then use this to conditionally compile either > of the two versions above. However, I have not tested them on the > different platforms. I've tested the current git tip on ARM (little-endian), x86_64, PowerPC, SPARC and S390x, and the tests pass everywhere. > > OK, so perhaps in this case the regression failure is to be expected, > > since the test is specifically designed to test long doubles... Would it > > be useful to distinguish full long double support from double-double > > support? > > Yes, I think so. The regression test is about a little-used trio > extension (the R flag for rounding) anyways, so it does not look > important. I am disabled it in git. Thanks! Regards, Stephen
signature.asc
Description: PGP signature