The argspecs of extension function/methods is broken in the Sphinx documentation

[hivert]

In the current Sphinx HTML doc, the extenstion function and methods have no arguments setup.

See for example the documentation of

reference/sage/symbolic/expression.html#sage.symbolic.expression.Expression.N

which was ok (see there) in Sage 4.8. The problem was introduced in sage-5.0.beta8.

We should add a regression test on that kinds of problem.

Apply:

• attachment: 12849_doctest.patch
• attachment: trac_12849-extfunc_argspec_html_fix-fh.patch

Component: documentation

Keywords: argspecs Cython

Author: Florent Hivert, Jeroen Demeyer

Reviewer: Mike Hansen

Merged: sage-5.0.beta14

Issue created by migration from https://trac.sagemath.org/ticket/12849

[jdemeyer]

Doctest for the issue

[jdemeyer]

comment:2

Attachment: 12849_doctest.patch.gz

I created a doctest for this issue, which obviously fails on recent Sage betas.

[hivert]

comment:3

Replying to @jdemeyer:

I created a doctest for this issue, which obviously fails on recent Sage betas.

Thanks ! I tried to figureout a way to call directly Sphinx but this is much easier. I'll try to work on this this afternoon. Do you have somewhere the various beta compiled so that I can rsync them on boxen to help bissecting ? So far the few guesses I made to find the culprit were wrong.

[jdemeyer]

comment:4

sage-5.0.beta5 is still okay.

[jdemeyer]

comment:5

Replying to @hivert:

Replying to @jdemeyer:

I created a doctest for this issue, which obviously fails on recent Sage betas.

Thanks ! I tried to figureout a way to call directly Sphinx but this is much easier. I'll try to work on this this afternoon. Do you have somewhere the various beta compiled so that I can rsync them on boxen to help bissecting ? So far the few guesses I made to find the culprit were wrong.

Look at http://boxen.math.washington.edu/home/release/. There should be binaries for all Sage betas, made on sage.math or boxen.math.

[hivert]

comment:7

It seems that I have a fix, but I don't understand how it worked before ! Still looking.

Florent

[jdemeyer]

comment:8

It looks like this was introduced in sage-5.0.beta8.

[jdemeyer]

comment:9

The culprit is #9128.

[hivert]

comment:10

Thanks for investigating.

Replying to @jdemeyer:

The culprit is #9128.

Strange ! It was an obvious candidate and I'm pretty sure I started by checking this one. Still I don't understand what could cause that in #9128. I'm testing my fix.

Florent

[hivert]

comment:12

Hi,

I got the fix ! I'm definitely the culprit. For strange reason the following lines were removed by #9128. Putting them back should fix the problem.

diff --git a/doc/common/conf.py b/doc/common/conf.py
--- a/doc/common/conf.py
+++ b/doc/common/conf.py
@@ -576,6 +576,8 @@ def find_sage_dangling_links(app, env, n
     newnode.append(contnode)
     return newnode
 
+from sage.misc.sageinspect import sage_getargspec
+autodoc_builtin_argspec = sage_getargspec
 
 def setup(app):
     app.connect('autodoc-process-docstring', process_docstring_cython)

[hivert]

Author: Florent Hivert, Jeroen Demeyer

[hivert]

comment:13

Attachment: trac_12849-extfunc_argspec_html_fix-fh.patch.gz

[hivert]

comment:14

Jeroen: I'm reviewing your patch. Can you review mine ?

Florent

[jdemeyer]

comment:15

Replying to @hivert:

Can you review mine ?

Sorry, no. I'm not at all familiar with the docbuilding process.

[mwhansen]

comment:16

Florent's patch is good, as is Jeroen (even if it is a bit sensitive to the particular HTML output used by Sphinx).

[mwhansen]

Reviewer: Mike Hansen

[hivert]

comment:17

Replying to @mwhansen:

Florent's patch is good, as is Jeroen (even if it is a bit sensitive to the particular HTML output used by Sphinx).

Thanks Mike !

[jdemeyer]

Merged: sage-5.0.beta14

[hughrthomas]

comment:19

Hi!

The doctest introduced by this patch, at line 22 of sage/misc/sagedoc.py, is causing the sagepad.org patchbot to fail on some (trivial, unrelated) patches. The problem seems to be that the doctest is trying to check the documentation when the documentation isn't built.

I don't know if this is really a problem with this patch, or with the patchbot, but I thought I would try reporting it here. If I should try elsewhere, please let me know!

You can see this failure at #12943, which is a completely trivial patch. (It's also making the sagepad.org patchbot fail on #10527. If you look at that ticket, though, don't look at the log for Volker's patchbot, which is having a different problem.)

cheers,

Hugh

[jhpalmieri]

comment:20

A complete set of documentation is an integral part of a Sage build, so doctests are supposed to fail if the documentation hasn't been built. This is not because of this ticket, I don't think.

[hughrthomas]

comment:21

Thanks. The issue seems to be with the particular patchbot -- I have posted to sage-devel about it.