Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Raise an error when calling make with MARCHes known to fail #11656

Merged
merged 1 commit into from
Jun 13, 2015
Merged

Conversation

nalimilan
Copy link
Member

Pre-SSE2 CPUs are not supported as some tests fail (#7185).
Better warn as soon as possible.

@yuyichao
Copy link
Contributor

The only issue I can see is that some distro maintainers might want to support older architectures (and for cases it works reasonably well...).

@nalimilan
Copy link
Member Author

Yeah, this was discussed on the other issue. For Fedora, I'm supposed to make i386 or i686-compatible builds. But in practice, better a Pentium4 build than nothing, and I don't think anybody would notice.

Shipping a build which doesn't pass the tests sounds like a very bad idea to me, especially if that's just to pretend supporting machines that nobody uses.

@yuyichao
Copy link
Contributor

Good enough for me. (I've also changed my ArchLinux julia-git package to pentium4 and claiming i686 arch)...

Just want to raise a concern I've seen from distro maintainers but if it has already been discussed I'll be totally find with it.

@tkelman
Copy link
Contributor

tkelman commented Jun 10, 2015

Actually we build 32-bit binaries right now with

MARCH = i686
JULIA_CPU_TARGET = pentium4

since there was a bug having to do with 32 bit calling convention in one of the dependencies, I think SuiteSparse.

@nalimilan
Copy link
Member Author

@tkelman Interesting. So JULIA_CPU_TARGET is what matters for this issue? I guess I should check for that variable instead, and suggest passing MARCH=i686 JULIA_CPU_TARGET=pentium4 OR MARCH=pentium4 if you don't build the dependencies. FWIW, the latter works fine in Fedora, and it would be a shame not to use SSE2 everywhere in Julia if we require it.

@tkelman
Copy link
Contributor

tkelman commented Jun 10, 2015

I agree but if we're going to adjust the buildbot configuration we should carefully test all 32-bit platforms on both release-0.3 and master first, to make sure it doesn't bring back any bugs.

@nalimilan
Copy link
Member Author

@tkelman Sorry, I had missed your comment. I wasn't really suggesting to change the buildbots right now -- just to give both suggestions to the user.

OK to merge the updated version?

@tkelman
Copy link
Contributor

tkelman commented Jun 13, 2015

I don't think I've built successfully with MARCH=i386, I typically use i686.

@nalimilan
Copy link
Member Author

OK, changed to i686.

@tkelman
Copy link
Contributor

tkelman commented Jun 13, 2015

I pushed a commit here to make sure this doesn't break 32 bit appveyor

@tkelman
Copy link
Contributor

tkelman commented Jun 13, 2015

my mistake, sorry, I already covered this in

echo "override JULIA_CPU_TARGET = pentium4" >> Make.user

Pre-SSE2 CPUs are not supported as some tests fail (#7185).
Better warn before the build.
tkelman added a commit that referenced this pull request Jun 13, 2015
Raise an error when calling make with MARCHes known to fail
@tkelman tkelman merged commit 4fe1f16 into master Jun 13, 2015
@tkelman tkelman deleted the nl/SSE2 branch June 13, 2015 17:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants