Cython interface to pselect() system call

[jdemeyer]

Useful for the new doctesting framework.

Depends on #14080
Depends on #12415

Component: misc

Author: Jeroen Demeyer

Reviewer: Volker Braun

Merged: sage-5.9.beta0

Issue created by migration from https://trac.sagemath.org/ticket/14079

[jdemeyer]

Dependencies: #14080

[jdemeyer]

comment:2

I'm intentionally not setting it to needs_review yet, because I want to first have a use case (that will be #12415). But that being said, the code here is supposed to be finished.

[roed314]

comment:3

Testing on 5.6 I get a doctest framework error (since you use the new ....: continuation). But testing on 5.7.beta1 with #12415 I get a timeout on line 308 (r, w, x, t = PSelecter().pselect([f,g,f,f,g])).

[jdemeyer]

comment:4

Replying to @roed314:

with #12415 I get a timeout on line 308 (r, w, x, t = PSelecter().pselect([f,g,f,f,g])).

On which kind of system is that?

[roed314]

comment:5

Replying to @jdemeyer:

Replying to @roed314:

with #12415 I get a timeout on line 308 (r, w, x, t = PSelecter().pselect([f,g,f,f,g])).

On which kind of system is that?

OS X 10.6.8, Core 2 Duo.

[jdemeyer]

comment:6

Updated patch (I haven't looked at the OS X specific issue).

[roed314]

comment:7

See http://stackoverflow.com/questions/14045801/different-pselect-behaviour-on-osx-vs-linux for a possible explanation of what's going on on OS X.

[jdemeyer]

comment:8

Thanks for the pointer, but that doesn't seem to be the problem: no signals are involved in the failing tests you mention.

[jdemeyer]

comment:9

New version, works for me on OS X 10.6 with the latest version of #12415. Note that parallel testing on OS X is broken, but normal testing (with -p 1) works.

[roed314]

comment:10

Apply failed....