Don't check for the exact zn_poly version

[timokau]

Currently the doctests check for the version of zn_poly. That makes the doctests unnecessarily brittle. If somebody updates zn_poly they have to make a change at an unreleated position and it is an annoyance for packaging.

Component: distribution

Author: Timo Kaufmann

Branch/Commit: 83b23c8

Reviewer: Jeroen Demeyer

Issue created by migration from https://trac.sagemath.org/ticket/25311

[timokau]

Author: Timo Kaufmann

[jdemeyer]

comment:3

Why patch this specific test? I mean, I think that most of misc/package.py is problematic for packagers.

[timokau]

comment:4

Thats the only one I had to patch. I generate the files in the installed dir at build time by iterating through sages dependencies. I only have to translate two package names.

zn_poly has version 0.9 in nixpkgs, without the patch part. I could translate that just as the names, but that would need to be fixed with every change to the sage version.

And even when not considering packaging I think its best not to test for the exact version. That will always break the doctests with every zn_poly version, even if there was no api break.

[jdemeyer]

comment:5

Replying to @timokau:

Thats the only one I had to patch. I generate the files in the installed dir at build time by iterating through sages dependencies. I only have to translate two package names.

Interesting. I don't know what other distros do.

And even when not considering packaging I think its best not to test for the exact version. That will always break the doctests with every zn_poly version, even if there was no api break.

I think we specifically chose zn_poly in this doctest because it is dead upstream (last release in 2008) and the last Sage-specific patch was in 2013. So the scenario that you describe is mostly hypothetical.

[timokau]

comment:6

Replying to @jdemeyer:

Interesting. I don't know what other distros do.

At least the arch linux package just fails the package.py tests.

I think we specifically chose zn_poly in this doctest because it is dead upstream (last release in 2008) and the last Sage-specific patch was in 2013. So the scenario that you describe is mostly hypothetical.

Ah, I see. I would still prefer not to rely on this. Does that test provide any value? That installed packages are detected correctly is tested in list_packages and that the version is split from the name correctly is tested in pkgname_split.

[jdemeyer]

comment:7

OK, fine. Could you just replace # not tested by # random?

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

e1526f8

Don't test for the exact zn_poly version

[sagetrac-git]

Changed commit from 67a011d to e1526f8

[timokau]

comment:10

Done

---

New commits:

e1526f8

Don't test for the exact zn_poly version

[jdemeyer]

comment:11

You are changing unrelated files

[sagetrac-git]

Changed commit from e1526f8 to 83b23c8

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

83b23c8

Don't test for the exact zn_poly version

[timokau]

comment:13

Sorry, no idea how that happened. I can't remember touching those files.

---

New commits:

83b23c8

Don't test for the exact zn_poly version

[jdemeyer]

Reviewer: Jeroen Demeyer

[vbraun]

Changed branch from u/gh-timokau/znpoly-version to 83b23c8