Mac OS binaries looking for libgfortran

[carljv]

Version 0.2.0-prerelease+3790 
Commit ee92383 2013-09-19 07:43:21 UTC
x86_64-apple-darwin12.4.0

I've installed the Mac OS binary beta version above and am running into an odd problem. When I call a function like Base.inv, I get the following error:

julia> inv([1. 2.; .5 20.])
ERROR: error compiling inv: error compiling inv: error compiling trtri!: could not 
load module liblapack: dlopen(liblapack.dylib, 1): Library not loaded: /usr/local
/Cellar/gfortran/4.8.1/gfortran/lib/libgfortran.3.dylib
Referenced from: /Applications/Julia-0.2.0-pre-ee92383fb8.app/Contents
/Resources/julia/lib/julia/liblapack.dylib
Reason: image not found 

Now, I have a gfortran compiler installed (via Homebrew, but a 4.7 version -- there's no directory for 4.8.1 like Julia seems to be looking for here). But I find it odd that the binary is looking for a compiler like this anyway. Am I missing something in the install?

Thanks,
-c.

[stevengj]

This is definitely a bug. The binary build should not have a dependency on /usr/local/Cellar. @ViralBShah?

[JeffBezanson]

Yes, definitely a bug, but it's not looking for a compiler, but the libgfortran runtime library.

cc @staticfloat

[carljv]

Right - it looks like the same type of problem as : #1703

[staticfloat]

This should be addressed in 22f7d4f, please redownload the latest binary and retry.

[carljv]

Seems to be working now, though when I call inv, I do get the following warnings:

julia> inv([2 1; .4 1.5])
Warning: Possible conflict in library symbol dtrtri_
Warning: Possible conflict in library symbol dgetri_
Warning: Possible conflict in library symbol dgetrf_
2x2 Array{Float64,2}:
  0.576923  -0.384615
 -0.153846   0.769231

Not sure if that's just a separate LAPACK-related problem. I can open a new issue for that if so.

[staticfloat]

Those are normal. We actually have an issue about suppressing those
warnings completely, I would find the issue but I'm on my phone at the
moment.
On Sep 23, 2013 9:05 AM, "carljv" [email protected] wrote:

Seems to be working now, though when I call inv, I do get the following
warnings:

julia> inv([2 1; .4 1.5])
Warning: Possible conflict in library symbol dtrtri_
Warning: Possible conflict in library symbol dgetri_
Warning: Possible conflict in library symbol dgetrf_
2x2 Array{Float64,2}:
0.576923 -0.384615
-0.153846 0.769231

Not sure if that's just a separate LAPACK-related problem. I can open a
new issue for that if so.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/4315#issuecomment-24931381
.

[stevengj]

We've had numerous issues with BLAS/LAPACK symbol conflicts in the past, e.g. #1642; it seems like this keeps popping up.

[staticfloat]

#1642 is a different issue. Same symptoms, but different underlying cause. The cause of this is that we're linking against Accelerate, and various libraries pull the BLAS symbols into the current namespace multiple times. Since they are the same symbols, the warnings are harmless.

#3342 is the issue I was thinking of that would perhaps give us more control over when those warnings are printed and when they aren't.

I'm glad it's working for you, @carljv! I'm going to close the issue now.