Define our own PariOUT

[jdemeyer]

Instead of changing defaultOut in sage/libs/pari/gen.pyx, we should instead define our own PariOUT structure and use that.

Component: interfaces

Keywords: pari

Author: Jeroen Demeyer

Reviewer: Robert Bradshaw, John Cremona

Merged: sage-4.7.1.alpha0

Issue created by migration from https://trac.sagemath.org/ticket/11230

[jdemeyer]

Attachment: 11230_pariout.patch.gz

[robertwb]

comment:2

LGTM, pending adding the trac number to the ticket description.

Why in the world did we define __x to do ctypedef __x pariout_t?

[jdemeyer]

comment:3

Replying to @robertwb:

LGTM, pending adding the trac number to the ticket description.

This is no longer necessary (see http://groups.google.com/group/sage-devel/browse_thread/thread/f5a9c012f6299a9e/a3c2af4a7700bca3)

Why in the world did we define __x to do ctypedef __x pariout_t?

I didn't write that code, but probably to work around some limitation/bug in an earlier version of Cython?

[jdemeyer]

Reviewer: Robert Bradshaw

[JohnCremona]

comment:4

I know this has a positive review from the omnipotent testbot, but when I apply this to a freshly-rolled 4.7.alpha5 and do a full test I get two failures in sage/misc/sagedoc.py:

File "/home/jec/sage-4.7.alpha5.pari/devel/sage-main/sage/misc/sagedoc.py", line 874:
    sage: len(search_doc('tree', interact=False).splitlines()) > 2500  # long time
Expected:
    True
Got:
    False
**********************************************************************
File "/home/jec/sage-4.7.alpha5.pari/devel/sage-main/sage/misc/sagedoc.py", line 489:
    sage: 'abvar/homology' in _search_src_or_doc('doc', 'homology', 'variety', interact=False)
Expected:
    True
Got:
    False

I built Sage with "make build" rather than "make", in case that is relevant. Meanwhile I switched the tag to "needs info" but will continue with this applied and go on to #11130.

[jdemeyer]

comment:5

Replying to @JohnCremona:

I built Sage with "make build" rather than "make", in case that is relevant.

Probably it is, because the doctest failures you mention are related to the documentation. So I would guess that make doc-html would solve those problems.

[JohnCremona]

comment:6

Replying to @jdemeyer:

Replying to @JohnCremona:

I built Sage with "make build" rather than "make", in case that is relevant.

Probably it is, because the doctest failures you mention are related to the documentation. So I would guess that make doc-html would solve those problems.

I guessed so, so am about to try again after sage -docbuild all html has finished...

[JohnCremona]

comment:7

Replying to @JohnCremona:

Replying to @jdemeyer:

Replying to @JohnCremona:

I built Sage with "make build" rather than "make", in case that is relevant.

Probably it is, because the doctest failures you mention are related to the documentation. So I would guess that make doc-html would solve those problems.

I guessed so, so am about to try again after sage -docbuild all html has finished...

As expected, the problem went away. I suppose it was silly to base the full test on devel/sage and not devel/sage/sage given that I had not built the docs.

[JohnCremona]

Changed reviewer from Robert Bradshaw to Robert Bradshaw, John Cremona

[jdemeyer]

Merged: sage-4.7.1.alpha0