Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

Loading…

Revert fortran? #19179

Closed
adamv opened this Issue · 24 comments

6 participants

@adamv
Owner

Fortran 4.8 seems to have caused a ton of issues to be opened; should this be reverted? Is anyone here a domain expert?

@samueljohn

Perhaps related from my tap: Homebrew/homebrew-python#26

As I could not reproduce it, my suspicion was that older Macs than core-i5 or i7 have difficulties. So perhaps it could also be an homebrew issue because of how we handle flags...
I have not further investigated this. I would not call myself a fortran expert, not even user.

@rajeeja

One sleepless night yesterday, tried gfortran 4.2, 4.7 and 4.8 with open-mpi, open-mpi builds fine, but I get run time errors in make check while building a software (MOAB). All of 'em failed. Unfortunately "--disable fortran" at this point.

P.S. I have a MacBook Pro late 2011 i7 processor

@rajeeja

Something else is messed up..In one for the open-mpi forums it said: mpicc vs mpiCC cause problems while linking with clang.
It's an old post and is codes that have both C and C++ code co existing face this problem.
http://www.open-mpi.org/community/lists/users/2008/04/5311.php

@samueljohn

I guess we are facing different problems here.
One the one hand older-than-sandybridge-Macs report problems with gfortran 4.8. I could reproduce this with scipy from my homebrew/python tap. However, even with gfortran 4.7, I get the same error.
So I wonder if this is really related to gfortran 4.8.

On the other hand you Problem @rajeeja, which seems open-mpi and perhaps clang related.

@manphiz
Collaborator

Open-mpi and Clang problem will be solved from 1.7+, which is the current feature release.

@samueljohn

@manphiz 1.7+ of what, open-mpi?

@manphiz
Collaborator
@RuiPereira

Hi. I managed to fix my gfortran 4.8 by reinstalling gmp 5.1.1 with --build-from-source instead of using the bottled version. The bottled gfortran works fine now, so it seems there's some problem with the gmp bottles on older systems. The fix worked on both dual-core 64-bit core2 and dual-core 64-bit penryn with 10.8.3.

@mikemcquaid mikemcquaid closed this issue from a commit
@mikemcquaid mikemcquaid gmp: remove bottle.
Closes #18944.
Closes #19179.
cb72b5f
@mikemcquaid
Owner

I've removed the gmp bottle for now. I want to try and work out why our generic CFLAGS for bottles aren't working properly though and I'll rebottle when I've worked this out.

@adamv
Owner

Should we close the other open "Fortran" tagged issues with instructions for rebuilding?

@mikemcquaid
Owner

Shouldn't need to rebuild now, just reinstall gmp.

@samueljohn
@brandt brandt referenced this issue from a commit
Commit has since been removed from the repository and is no longer available.
@mikemcquaid mikemcquaid referenced this issue from a commit
Commit has since been removed from the repository and is no longer available.
@mikemcquaid mikemcquaid referenced this issue from a commit
Commit has since been removed from the repository and is no longer available.
@mikemcquaid mikemcquaid referenced this issue from a commit
Commit has since been removed from the repository and is no longer available.
@mikemcquaid mikemcquaid referenced this issue from a commit in mikemcquaid/homebrew
@mikemcquaid mikemcquaid Fix --build-bottle CFLAGS.
The CFLAGS were previously not generic enough.

References #18944.
References #19179.
646150a
@mikemcquaid
Owner

I've fixed the bottling cflags for Clang, rebuilt the gmp bottle and pushed the new bottle in bd1cdb4. @RuiPereira, @samueljohn and others: can you see if it works properly for you now? Thanks!

@RuiPereira

@mikemcquaid : still failing for me...

$ brew info gmp
gmp: stable 5.1.1 (bottled)
http://gmplib.org/
/usr/local/Cellar/gmp/5.1.1 (12 files, 2.5M) *
  Poured from bottle
https://github.com/mxcl/homebrew/commits/master/Library/Formula/gmp.rb

$ git log --pretty=oneline -n 1 /usr/local/Library/Formula/gmp.rb
bd1cdb4c38b258a330c5dd415ca23abc656281a0 gmp: re-add (hopefully) fixed bottles.

$ brew install salt
==> Downloading http://supernovae.in2p3.fr/~guy/salt/download/snfit-2.2.2b.tar.gz
Already downloaded: /Library/Caches/Homebrew/salt-2.2.2b.tar.gz
==> Using Homebrew-provided fortran compiler.
This may be changed by setting the FC environment variable.
==> ./configure --prefix=/usr/local/Cellar/salt/2.2.2b --enable-gfortran
==> make install
f951: internal compiler error: Abort trap: 6
gfortran: internal compiler error: Abort trap: 6 (program f951)
../libtool: line 1086: 51415 Abort trap: 6           /usr/local/bin/gfortran -DHAVE_CONFIG_H -I. -I. -I.. -g -Wall -O3 -g -O3 -c dfact.F -fno-common -o .libs/dfact.o
make[1]: *** [dfact.lo] Error 1
make: *** [install-recursive] Error 1

$ brew --config
HOMEBREW_VERSION: 0.9.4
ORIGIN: git://github.com/mxcl/homebrew.git
HEAD: 097fd8f48fce6ced58797a4477207b2de2b40fe0
HOMEBREW_PREFIX: /usr/local
HOMEBREW_CELLAR: /usr/local/Cellar
CPU: dual-core 64-bit penryn
OS X: 10.8.3-x86_64
Xcode: 4.6.2
CLT: 4.6.0.0.1.1365549073
LLVM-GCC: build 2336
Clang: 4.2 build 425
X11: 2.7.4 => /opt/X11
System Ruby: 1.8.7-358
Perl: /usr/bin/perl
Python: /usr/bin/python
Ruby: /usr/bin/ruby => /System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/bin/ruby

also failing with a similar installation on a CPU: dual-core 64-bit core2

@mikemcquaid mikemcquaid referenced this issue from a commit
@mikemcquaid mikemcquaid gmp: remove bottles again.
Still getting "invalid instruction" errors.

References #19179.
Closes #19788.
c2cc581
@lovelace lovelace referenced this issue from a commit
@mikemcquaid mikemcquaid Fix --build-bottle CFLAGS.
The CFLAGS were previously not generic enough.

References #18944.
References #19179.
5e72dfd
@lovelace lovelace referenced this issue from a commit
@mikemcquaid mikemcquaid gmp: remove bottles again.
Still getting "invalid instruction" errors.

References #19179.
Closes #19788.
099d5aa
@samueljohn

@RuiPereira still failing for you?
I can use the gfortran bottle and then brew install salt. Builds nicely.
However, the brew test salt fails ... perhaps because I didn't install --with-data...

@RuiPereira

@samueljohn : yes it works fine with the current non-bottled gmp. The gfortran bottle was never the problem IMHO. The salt test depends on --with-data yes...

@samueljohn

Okay. If brew install gmp and it built from source, so this issue is solved until the bottles for gmp are added back. We have to keep that in mind.

I just wanted to give feedback and check if working right now.

@samueljohn

The salt test has to be fixed otherwise the bot.brew.sh will complain...

@RuiPereira

I'll work on it as soon as possible.

@samueljohn

Either we always install the data (how big are they?) or we only install a little data file for testing purpose. I would love to have the test up and running on the bots so we can always check for failures.

@RuiPereira

Data is ~ 13MB. I would vote to remove that option and just install the data by default, since that the code (and hence the test) is useless without it.

@samueljohn

Not sure. How much of the data is needed for the test?

@RuiPereira

Could shave a couple of MB, but the bulk of the data (~11MB: the template model used to fit user data) is needed for the test.

@mikemcquaid
Owner

13MB is nothing. Go for it.

@handyman5 handyman5 referenced this issue from a commit
Commit has since been removed from the repository and is no longer available.
@handyman5 handyman5 referenced this issue from a commit
Commit has since been removed from the repository and is no longer available.
@handyman5 handyman5 referenced this issue from a commit
Commit has since been removed from the repository and is no longer available.
@draftycode draftycode referenced this issue from a commit in draftycode/homebrew
@mikemcquaid mikemcquaid gmp: remove bottle.
Closes #18944.
Closes #19179.
8f19266
@draftycode draftycode referenced this issue from a commit in draftycode/homebrew
@mikemcquaid mikemcquaid Fix --build-bottle CFLAGS.
The CFLAGS were previously not generic enough.

References #18944.
References #19179.
15244ac
@draftycode draftycode referenced this issue from a commit in draftycode/homebrew
@mikemcquaid mikemcquaid gmp: remove bottles again.
Still getting "invalid instruction" errors.

References #19179.
Closes #19788.
8ffa617
@shelhamer shelhamer referenced this issue from a commit
@mikemcquaid mikemcquaid gmp: remove bottle.
Closes #18944.
Closes #19179.
b7da729
@shelhamer shelhamer referenced this issue from a commit
@mikemcquaid mikemcquaid Fix --build-bottle CFLAGS.
The CFLAGS were previously not generic enough.

References #18944.
References #19179.
93fa6bc
@shelhamer shelhamer referenced this issue from a commit
@mikemcquaid mikemcquaid gmp: remove bottles again.
Still getting "invalid instruction" errors.

References #19179.
Closes #19788.
ecd361c
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.