On Wed, Jun 25, 2014 at 01:41:02AM +0200, FX wrote: > > If I remove the previously installed gcc, the failure again occurs. > > So, it looks like the testsuite is picking up installed *.mod files > > over the freshly built *.mod files. This is not a showstopper. > > And this is not all the testsuite, but only ieee_1.F90. After a long > chase, it turns out it?s because it has a dg-options, which overrides > my IEEE options in ieee.exp. This is because I used dg-options when I > should have used dg-additional-options (phew!).
Testcase now passes. > > > I also read through the config/fpu-sysv.h diff. It looks correct > > to me. I do note that I suspect that there is a problem in > > config/fpu-sysv.h. > > I concur with your diagnosis. I?ve fixed this, and full new patch > is attached (including regenerated files). > > > Both Janne and Tobias have stated that they took a quick glance > > over the patch, and with my testing I think you should commit. > > Although you'll probably need to deal with other odd architectures. > > I?ll wait a few more days to commit, so others can comment/review > and I am sure to be around if there is fallout. > I tested the new diff and ieee_1.F90 on x86_64-*-freebsd. 'gmake check-gfortran' passed without any new issues. Thanks for taking on the IEEE features of Fortran. -- Steve