singular.version() has no doctest

[simon-king-jena]

Neither singular.version nor singular_version have doc tests.

Upstream: Completely fixed; Fix reported upstream

CC: @kedlaya

Component: interfaces

Keywords: singular version help.cnf

Author: Jori Mäntysalo

Branch/Commit: 87c73ac

Reviewer: Jeroen Demeyer

Issue created by migration from https://trac.sagemath.org/ticket/5994

[simon-king-jena]

comment:1

Remark:

Apparently the problem is on the side of Singular.

In Singular, freshly started, when you do

> system("--version");

there is the error (it says "? cannot open help.cnf").
If you repeat, no error.

If people think that my suggestion would break code, I suggest to have a new method singular.version_number().

[simon-king-jena]

comment:2

Replying to @simon-king-jena:

Apparently the problem is on the side of Singular.

And meanwhile it is reported upstream.

I do not know, however, if the problem still occurs with Singular 3-1-0 (I only have a beta version, and this still shows the error).

[simon-king-jena]

comment:3

Replying to @simon-king-jena:

Replying to @simon-king-jena:
And meanwhile it is reported upstream.

I do not know, however, if the problem still occurs with Singular 3-1-0 (I only have a beta version, and this still shows the error).

"Upstream" answered, and it seems that the problem is fixed in the official release.

I could verify that on the server in Oberwolfach, the error does not occur in Singular 3-1-0, but it does occur in Oberwolfach in Singular 3-0-4 (as being part of Sage). Opinions?

But there still remains the question if singular.version() ought to return a tuple of integers, rather than a lengthy string with all build informations.

[aghitza]

comment:4

Replying to @simon-king-jena:

"Upstream" answered, and it seems that the problem is fixed in the official release.

I could verify that on the server in Oberwolfach, the error does not occur in Singular 3-1-0, but it does occur in Oberwolfach in Singular 3-0-4 (as being part of Sage). Opinions?

We can probably wait until after Sage Days 15 for this?

But there still remains the question if singular.version() ought to return a tuple of integers, rather than a lengthy string with all build informations.

My personal preference would be for it to return a tuple of integers, but we could give it an optional argument with_build_info=True (or verbatim=True or something similar) to make it return the lengthy string.

[simon-king-jena]

comment:5

Replying to @simon-king-jena:

I could verify that on the server in Oberwolfach, the error does not occur in Singular 3-1-0, but it does occur in Oberwolfach in Singular 3-0-4 (as being part of Sage). Opinions?

The "Opinions?" bit should be two lines lower. Sorry.

It turned out that the error occurs in any Sage-built Singular version that I know: Singular 3-0-4 on sage.math, on two of my computers, in Oberwolfach, and Singular 3-1-0-Beta on sage.math and on one of my computers.

If Singular is not built by Sage, then the error apparently does not occur: With Singular 3-0-3 (very old) on one of my machines, Singular-3-1-0-Beta on my machine (same sources, modulo the usual patches, as the sage-built version!), and Singular-3-1-0 official release in Oberwolfach.

So, after all, it seems to me that it is all Sage's fault.

But there still remains the question if singular.version() ought to return a tuple of integers, rather than a lengthy string with all build informations.

Here is where I want to know some "Opinions"...

[williamstein]

comment:6

The attached patch fixes the output format for version to be more consistent with the other interfaces, e.g., gp.version(). It also programs around the issue with help files, which is not fixed in Singular-3-1-0... and I'm not at all clear why it is considered a bug in Singular by the people in the thread above.

[simon-king-jena]

comment:7

Attachment: trac_5994.patch.gz

Replying to @williamstein:

It also programs around the issue with help files, which is not fixed in Singular-3-1-0... and I'm not at all clear why it is considered a bug in Singular by the people in the thread above.

Note that I originally thought that the issue with help files is a problem of Singular, and clearly a bug: I mean, you ask for the version number and get an error; you ask again, and it works! It had reported it upstream.

But then the impression came across (see my last post) that it only occurs in Singular if it is built by Sage. In this case, it could be a problem with the patched version in Sage, which might be worth another ticket.

Cheers,
Simon

[williamstein]

comment:8

I think the first time you ask for the version, singular fires up its help system, and reports a bug about it not being properly configured by us for such use. Then it doesn't report that again, since it already did. I think it is very sensible behavior by Singular.

So can you review this?

[wjp]

comment:9

For the record, adding an empty $SAGE_ROOT/local/share/singular/help.cnf suppresses the error. There is actually a help.cnf file in the singular sources (, but it doesn't seem to get installed. Its contents don't look to be particularly relevant for sage at first glance, though.