I've identified the problem. fixup-libgfortran was sad because all the
consumers of libgfortran linked to lib{gfortran,quadmath,gcc_s} via the
"public" paths, which look like /usr/local/lib/gcc/4.9/xxx, whereas
libgfortran internally linked to libquadmath via a "private" path, which
looks like /usr/local/Cellar/gcc/4.9.1_2/lib/xxx. I've since fixed up
fixup-libgfortran.sh and uploaded a new binary.
-E
On Fri, Dec 26, 2014 at 9:20 AM, Mike Innes <[email protected]> wrote:
> If it helps at all, I saw that exact issue a few days back, on 0.3 and
> 0.4. I had to delete and rebuild the openblas included in the Julia repo. I
> don't know exactly how the Julia 0.3.x builds work but a stale openblas may
> be causing issues.
>
> On 26 December 2014 at 16:13, Tony Kelman <[email protected]> wrote:
>>
>> Hm, strange. We should be including libquadmath.0.dylib in the mac
>> binaries, making them self-contained is the goal and we have some
>> libgfortran-related code that should be working. That error is definitely a
>> homebrew-ism, do you know whether you already had a libopenblas.dylib on
>> your path when you were getting the error? I'm not sure what order the
>> library paths get searched on mac, I would hope we search our bundled
>> libraries before whatever happens to be on the user's system. But looking
>> at the code for blas_vendor() in base/util.jl, we might be picking up
>> system openblas first.
>>
>>
>> On Friday, December 26, 2014 6:44:30 AM UTC-8, Ed Scheinerman wrote:
>>>
>>> I had a (fixable) problem installing on my Macbook Air. Upon first
>>> launch in a terminal window I got this error:
>>>
>>> Warning: error initializing module LinAlg:
>>>
>>> ErrorException("error compiling __init__: error compiling check_blas:
>>> error compiling openblas_get_config: could not load module libopenblas:
>>> dlopen(libopenblas.dylib, 1): Library not loaded:
>>> /usr/local/Cellar/gcc/4.9.2_1/lib/gcc/4.9/libquadmath.0.dylib
>>>
>>> Referenced from: /Applications/Julia-0.3.4.app/
>>> Contents/Resources/julia/lib/julia//libgfortran.3.dylib
>>>
>>> Reason: image not found")
>>>
>>>
>>> Julia then started up but (not surprisingly) it wouldn't invert a
>>> matrix. The reference to "cellar" suggested a homebrew fix so I did
>>>
>>> $ brew install homebrew/science/openblas
>>>
>>> and relaunched Julia. All seems fine now. My Julia 0.3.3 worked fine. I
>>> think making the distribution self-contained would be preferable. Or might
>>> there be something odd with my setup?
>>>
>>>
>>>
>>>
>>>
>>> On Friday, December 26, 2014 6:43:42 AM UTC-5, Elliot Saba wrote:
>>>>
>>>> Hello all! The latest bugfix release of the 0.3.X Julia line has been
>>>> released. Binaries are available from the usual place
>>>> <http://julialang.org/downloads/>, and as is typical with such things,
>>>> please report all issues to either the issue tracker
>>>> <https://github.com/JuliaLang/julia/issues>, or email this list.
>>>>
>>>> As this is a bugfix release, there are not too many new big-item
>>>> features to announce, but if you are interested in the bugs fixed since
>>>> 0.3.3, this commit log
>>>> <https://github.com/JuliaLang/julia/compare/v0.3.3...v0.3.4> should
>>>> give you an idea of the effort put in by our team of backporters, as they
>>>> faithfully toiled to bring you an extremely early 2015 Christmas present.
>>>>
>>>> This is a recommended upgrade for anyone using any of the previous
>>>> 0.3.x releases, and should act as a drop-in replacement for any of the
>>>> 0.3.x line. We would like to get feedback if someone has a correctly
>>>> working program that doesn't work after this upgrade.
>>>>
>>>> Happy Hacking,
>>>> -E
>>>>
>>>
