[Debian-salsa-ci] More Salsa CI stats

[Aayush Raj]

Hi Otto!
> > Lucas just published a new version of trends.debian.net, and now it
> > also includes:
> >
> > https://trends.debian.net/#continuous-integration-on-salsa
> >
> > Now with both this and salsa-stats.debian.net we could perhaps plan
> > some campaign to drive adoption (activate salsa-ci.yml in all
> > packages)

This is very good insight! While the idea of activating salsa-ci.yml
in all packages is very cool, I am a bit concerned about the extra
resource consumption, especially from packages that don't actively
monitor or care about the results and failing pipelines.

Estimating from the overload we had a couple of weeks ago, Salsa
typically operates at around 15% of its total CI capacity (220 out of
1400 pipelines per day). So, I imagine that if we were to enable Salsa
CI sitewide today, the load would be manageable. (I’m intentionally
being conservative with the numbers here, the real value could be
somewhere around ~10% of capacity).

I'm neutral on the campaign itself, but I'm definitely interested in
exploring ways to raise awareness about Salsa CI and its benefits
among those who are currently unfamiliar with it. I like the idea of
sending a global broadcast message.
https://docs.gitlab.com/administration/broadcast_messages

---

Hi Andrea,

> - For the Salsa CI one, only checking for yaml files in source packages
>   leaves out a huge number of packages which used to follow the team's
>   reccomendations before
>   <https://salsa.debian.org/salsa-ci-team/pipeline/-/commit/112bee8323519d77c62105315635bb50815c1f79>,
>   which is less than one year ago (!).

Yes, I agree that we aren't certain how many projects still use
`recipes/debian.yml at salsa-ci-team/pipeline` in the CI config path.
Based on salsa-status.d.n, that percentage is approximately 17% (700
out of 4000). Note that this figure excludes dormant projects (the
other 9000) that have not yet been reflected on salsa-status.d.n. I
believe Salsa Admins could provide the exact numbers, but I'm unsure.

Regardless, I think the stats on trends.d.n are very useful for
getting a rough estimate of global adoption and its increase. I can
see that **at least** 13773 projects use/have used Salsa CI. Also, an
FYI, from Otto's
https://optimizedbyotto.com/post/debian-salsa-ci-gsoc-2025: See the
url's grep result
https://codesearch.debian.net/search?q=salsa+path%3Adebian%2F.*.yml&literal=0
(it has some false positives though)

>   Also, a salsa.ci.yml file by
>   itself doesn't guarantee that CI is being used, as one has to still
>   open up GitLab's Web UI config pages and tell the service that yes,
>   that file should be used to run CI.

I think anyone who adds a CI config in the `debian/` directory would
also enable it, and subsequent forks would inherit those settings.
There might be some who disable CI later, but that number should be
small. That's just my assumption though, could be wrong =)


Hope the numbers helped!
Good day!


On Wed, 17 Sept 2025 at 12:59, Andrea Pappacoda <tachi at debian.org> wrote:
>
> Hi Otto,
>
> On Tue Sep 16, 2025 at 8:22 AM CEST, Otto Kekäläinen wrote:
> > Lucas just published a new version of trends.debian.net, and now it
> > also includes:
> >
> > https://trends.debian.net/#continuous-integration-on-salsa
>
> In priciple, this is great! But in practice, I'm not sure how these
> stats are useful in their current state:
>
> - For the Salsa CI one, only checking for yaml files in source packages
>   leaves out a huge number of packages which used to follow the team's
>   reccomendations before
>   <https://salsa.debian.org/salsa-ci-team/pipeline/-/commit/112bee8323519d77c62105315635bb50815c1f79>,
>   which is less than one year ago (!). Also, a salsa.ci.yml file by
>   itself doesn't guarantee that CI is being used, as one has to still
>   open up GitLab's Web UI config pages and tell the service that yes,
>   that file should be used to run CI.
>   The only case in which the check as implemented would be useful is the
>   future and hypothetical case where Salsa would automatically run CI
>   upon finding a salsa-ci.yml file in the debian directory, which isn't
>   the case now.
>   Until then, I find this check not that meaningful.
> - As mentioned already in the patch adding the lintian classification
>   tag for gbp.conf files, saying "the package uses gbp because a conf
>   file exists" is overly optimistic. I personally always add a gbp file
>   to my packages because I find it nice to write somewhere what my
>   branch layout is, or whether this package uses pristine-tar, etc. The
>   only gbp component I personally use is gbp-dch, and that doesn't even
>   require a conf file for what I use it for. Should I start deleting
>   those conf files, and add that information in d/README.source?
>
> Does this make sense? Let me know :)
>
> Bye!
>
> --
> Debian-salsa-ci mailing list
> Debian-salsa-ci at alioth-lists.debian.net
> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-salsa-ci