GMP installation issues with 3.0

[jeanluct]

Margaux is having some issues with the Mac binaries. Basically, they don't work out of the box if you don't have GMP installed. What's the easiest way to resolve this?

Two easiest things to do:

• Provide _noGMP_ binaries for OSX.
• For Linux, add to the guide instructions on how to install GMP. Something like sudo apt-get install libgmp.

[mbudisic]

I believe both can be done for both distributions, no? In certain cases, users even on linux will not have sufficient privileges to install GMP using apt, and I wouldn't want to support manual installation of GMP for them.

On Mac, "homebrew" for now does a decent job of replicating the ease of use of apt. We could add that to "install GMP" suggestions.

[jeanluct]

Ok, well let's test it with Margaux, see if it works!

[jeanluct]

Update: Marko compiled _noGMP_ binaries and they work for Margaux.

New problem: -DBRAIDLAB_USE_GMP=0 does not work on Mac OSX. Linux uses GNU make, and Mac must use BSD make or something.

[mbudisic]

No no. That flag works fine. It's the specific syntax
$ make BRAIDLAB_USE_GMP=0
vs
$ make BRAIDLAB_USE_GMP=1
that does not work.

[jeanluct]

Ah, right. No need for -D. But it work on Linux.

[mbudisic]

This works on Mac OS X:
$ BRAIDLAB_USE_GMP=1; make
vs
$ BRAIDLAB_USE_GMP=0; make
but not on linux.
I would guess this is due to Mac OS X using bash 3.2 and linux 4.2 on my machines. make is actually the same exact version on both of the systems I'm using.

[jeanluct]

Does this work on both?
$ export BRAIDLAB_USE_GMP=1; make
vs
$ export BRAIDLAB_USE_GMP=0; make

[mbudisic]

OK, I figured out what it is.
the syntax
$ make BRAIDLAB_USE_GMP=0
is overriden by the global variable. In other words, first setting a global variable prevents make from using the variable passed as an argument. Global has to be cleared first before this syntax works. (And then it works on both Mac and linux, regardless of bash version)

[jeanluct]

So this wasn't working for you because you had a global defined? So for most people it won't be a problem?

[mbudisic]

Yes, I believe so. I had a global leftover from compiling makedist.

[jeanluct]

Ok, so it's not a bug. Just an unfortunate convention...

On Thu, Dec 11, 2014 at 1:52 PM, Marko Budisic notifications@github.com
wrote:

Yes, I believe so. I had a global leftover from compiling makedist.

—
Reply to this email directly or view it on GitHub
#101 (comment).

[jeanluct]

So does anything really need to change? Should we clarify something in the manual?

[mbudisic]

Not really, only if you feel like make BRAIDLAB_USE_GMP=0 syntax being overloaded by environ variable is something to note.

[jeanluct]

TODO: Marko writes a short blurb in guide about using homebrew to install GMP on OSX.

Then we can release 3.0.1.