Mathematica interface is broken by Mathematica 9

[EmmanuelCharpentier]

Something seems to have changed in Mathematica version 9 interface with "the rest of the world".

Setup(s) : Debian wheezy with self-compiled sage v 5.4.1 then v 5.5, Mathematica Linux 64 bits V8 then V9.

(1) sage v 5.4 <--> Mathematica V8 : OK

(2)sage v 5.5 <--> Mathematica V9 : doesn't work. Sage reports to be "unable to start Mathematica". However, Mathematica works both from the command line (math) or from the GUI (mathematica).

(3) sage v 5.5 <--> Mathematica V8 : OK again.

I've also seen (1) and (3) on a smallish 32 bit machine (Again, debian wheezy + self-compiled sage (this was slooow..)).

This issue has been discussed on sage-support ; William Stein suggested writing an auxilliary app communicating with Mathematica via the MathLink protocol.

Component: interfaces: optional

Reviewer: Dima Pasechnik

Issue created by migration from https://trac.sagemath.org/ticket/13892

[NathanDunfield]

comment:1

I have replicated this with Sage 5.10 and Mathematica 9 on OS X (Leopard). I believe the problem is that Mathematica 9 quits as soon as it hits an EOF if you redirect its output:

 math9 > /tmp/math9.log   # quits immediately after printing the banner to the logfile. 

 math8 > /tmp/math8.log   # Takes keyboard input and prints output to the logfile as expected.

Here's one workaround for this I found involving "tail -f" and a FIFO:

http://mathematica.stackexchange.com/questions/23217/keep-mathkernel-running-in-background-speed-up-execution-time

but perhaps there's some easy way to imitate this just using pexpect?

[mkoeppe]

comment:7

outdated, should close

[dimpase]

Reviewer: Dima Pasechnik