Comma in upstream version causes error.

simplert has a version "1,0" with a comma inside: https://github.com/vvviperrr/SimpleRT/releases running f-droid checkupdates gives the following error:

WARNING: ...com.viper.simplert : VCS error while scanning app com.viper.simplert: Git checkout of '1' failed
==== detail begin ====
error: pathspec '1' did not match any file(s) known to git.
==== detail end ====
INFO: Finished.

Should UCM for that app be disabled (until this bug is resolved)?

Milestone changed to %0.7

I can reproduce this in the latest version in git. This is a classic issue where using python-git would help us. All the parsing edge cases would be handled already. I think I can fix this now though.

So there are a few issues here:

• The tags are currently fetched by parsing the output of git log --tags --simplify-by-decoration --pretty=format:%d which uses commas as part of the data format: (HEAD, tag: foo, tag: bar, origin/master, origin/HEAD, master)
• the .txt metadata format uses comma to separate versionName from versionCode: Build:1.0,1
• that means commas are banned here: https://gitlab.com/fdroid/fdroidserver/blob/329e0247d5b4fae6c0e1ac25ffd784ce851d54b1/fdroidserver/common.py#L713

To fix this, we'd need to parse the tags using python-git, then fix the metadata.txt format to handle commas in the versionName. That could be done by including all commas in the versionName except the last one. versionCode will never have commas since it must be an integer.

That could be done by including all commas in the versionName except the last one. versionCode will never have commas since it must be an integer.

sounds good.

To fix this, we'd need to parse the tags using python-git

I hope you're not suggesting this for 0.7.

Milestone removed

Milestone changed to %1.0

One little wrench in my plan: vcs.latesttags() needs to sort the tags in date order, and python-git does not seem to do that.

bug report to upstream: https://github.com/vvviperrr/SimpleRT/issues/9

We need sorting by date because otherwise alphabetical order will not do. We can either use the latest N tags, or all the tags. And the latter is very, very slow for large repos with tons of tags.