Re: Compilation failure for openblas-0.3.7-1 under OSX 10.11.6

"Homeier, Derek" <[email protected]>
Newsgroups gmane.os.apple.fink.beginners
Message-ID <[email protected]>
Hi David,
> 
>  I have not so far been able to find an update for CLT to reach 8.2.1; the only version I see on the Apple Developer site that is listed for use with OSX 10.11 and Xcode 8.2 turns out after re-installation to be the version I already had (8.2.0.0.1.1480973914).  For what it's worth I did see the same version mismatch showing up in at least one message on the fink-beginners archive (https://sourceforge.net/p/fink/mailman/message/35611744), where as far as I could see the mismatch was not related to the problem (but I could well be wrong about that).  I tried editing the install script as you requested even though CLT was not updated (I inserted the new lines immediately after the "if" block that starts "if [ "$darwin_vers" -ge 18 ]"), but that gave the same failure on attempting to compile openblas as before.  If you can guide me to a proper update for CLT I will of course be happy to try again!  Otherwise it's maybe starting to look like restriction might have to be the answer.

I’ve just noticed that version 0.3.8 is out and am testing building that - it is unlikely
to get us rid of this problem, but maybe I can find some hints how to work around it.
But did I understand you correctly that the 0.3.6 build succeeded on your 10.11 machine?
That’s indeed odd, since the entire smallscaling.c with the clock_gettime() call has not
changed in 4 years. That would really hint to either changes in the compilation options
I made between 0.3.6 and 0.3.7, or indeed your mismatch in the CLT version.
But I also remember having similar version mismatches, when there simple was no
newer version of the CLTs available from Apple…

One more thing I also changed in 0.3.7 was switching to gcc-fsf-9 for the C compilation -
that because the FSF gfortran is anyway used for the Fortran part, and a few tests were
giving better results with the FSF C compiler, but generally the compilation with Apple’s
clang worked just as well. So you might try to switch back to that by simply setting

SetCC: clang

(might seem counterintuitive since it is the CLT clang having the version mismatch,
but it could be worth just giving a try).

HTH
				Derek


_______________________________________________
Fink-beginners mailing list
[email protected]
List archive:
https://sourceforge.net/p/fink/mailman/fink-users/
Subscription management:
https://lists.sourceforge.net/lists/listinfo/fink-beginners
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.