On Fri, Jul 06, 2018 at 06:38:55AM -0500, Segher Boessenkool wrote: > On Fri, Jul 06, 2018 at 01:51:37AM -0400, Michael Meissner wrote: > > case "$target:$with_long_double_format" in > > > - xpowerpc64*-*-linux*:*) > > So this case could never happen. The changelog should mention it fixes > that bug (and having it as a separate patch is much preferred!)
I assume what happened is I accidently added the 'x' to the working copy after submitting the patch, but before committing it and I didn't notice it. Since it is in configuration support, it isn't part of the test sutie, and it wasn't noticed. I can add a line to the ChangeLog if desired. > Other than this thing, the original code was easier to read. What does > this part of the patch improve? You complained that you were getting errors when using the system glibc (based on 2.27 on an Ubuntu system) and using --with-long-double-format=ieee (where it would die in the middle of building libstdc++-v3). I wrote the patch to check that the glibc has the support so it fails when configuring the compiler and gives a sensible message (need glibc 2.28). But it doesn't really change anything. If you have an appropriate glibc, it will build without the patch, and if you don't, it will fail. But it should be friendlier to people building the compiler to understand why it failed. I could duplicate the tests for glibc 2.28 (and AT-next alpha) for big endian and little endian if desired, but it seemed clearer to me to combine the code rather than duplicate the tests. -- Michael Meissner, IBM IBM, M/S 2506R, 550 King Street, Littleton, MA 01460-6245, USA email: meiss...@linux.ibm.com, phone: +1 (978) 899-4797
