Hi, my graph theory development is CI testing against mathcomp+finmap and I'm getting strange build errors (not failures) because coq-mathcomp-finmap-1.4.0
fails to build in the CI environment while it builds fine locally. Does someone have an idea what may be going on? I hope this is the the right place to ask. See the build info.
@Christian Doczkal Are you on Coq 8.11.1?
finmap-1.4.0
is not compatible with the recently released Coq 8.11.1
Yes, you are right, I am still on 8.11.0
:cry:
@Anton Trunov do you know what is the source of the incompatibility?
@Cyril Cohen it’s this https://github.com/math-comp/finmap/issues/62
it looks like that change from coq.dev finally made its way into coq 8.11.1
I opened https://github.com/coq/opam-coq-archive/pull/1231 to constraint finmap to 8.11.0 at most
I guess it could be a time to make a new release. I tried locally and finmap’s master branch is compatible with coq 8.11.1
I concur, finmap.dev does build agains coq.dev and mathcomp.dev.
So if I release finmap-1.4.1 using the current master branch, everything will be alright?
that’s right
I relaunched the CI on relevant docker images just to be sure
Here is a test on my machine (macOS)
$ opam install coq-mathcomp-finmap.dev
The following actions will be performed:
∗ install coq-mathcomp-finmap dev
<><> Gathering sources ><><><><><><><><><><><><><><><><><><><><><><><><><><> 🐫
[coq-mathcomp-finmap.dev] synchronised from git+https://github.com/math-comp/finmap.git#master
<><> Processing actions <><><><><><><><><><><><><><><><><><><><><><><><><><> 🐫
∗ installed coq-mathcomp-finmap.dev
Done.
$ coqc -v
The Coq Proof Assistant, version 8.11.1 (April 2020)
compiled on Apr 6 2020 14:51:38 with OCaml 4.09.0
will do then...
Great! If you push a tag, I can take care of the opam repository
great
here you go https://github.com/math-comp/finmap/releases/tag/1.4.1
Thanks! Here is the PR: https://github.com/coq/opam-coq-archive/pull/1232
Thanks @Anton Trunov and @Cyril Cohen for the quick reaction. This way I can hopefully have my CI green in the very near future :+1:
Thanks everyone (and no problem)! Once the PR is merged it’s usually a matter of a few ours before the packages are available on opam.
@Cyril Cohen Perhaps we should add finmap to Coq’s CI to avoid this breakage in the future. I can help with that too. Although, as you may already know there are certain conditions described here https://github.com/coq/coq/blob/master/dev/ci/README-users.md
I'm a bit uneasy with 8.11.1 introducing this kind of incompatibility
@Emilio Jesús Gallego Arias me too
so feel free to propose a revert for 8.11.2 if you think that would be the most appropiate
I fixed all the dependencies of coq-mathcomp-finmap, but I’m not sure what else might be broken
we'll see, just mentioning it as to encourage the revert if it is causing too much pain; certainly breaking finmap was not intended
regarding inclusion of projects in the CI, including finmap
Ilya's new stuff etc...
the policy is that we welcome anything upstream wants to include
we are gonna get less pro-active in adding stuff to the CI ourselves, so in some sense, if the developers of contribs don't bother submitting to Coq's CI , then we won't add them
we are just waiting for the process to greatly improve
Is it fine to add ceramist to Coq’s CI provided it compiles for about 50 minutes on 2 cores?
Sure it is
we have far more heavy stuff
Awesome! Let us wait for Cyril’s decision on finmap first.
I added issue https://github.com/math-comp/finmap/issues/65 to track this.
Great, thanks @Anton Trunov
My pleasure :)
@Anton Trunov you know why coq-ceramist is called ceramist ?
That's an interesting name
@Emilio Jesús Gallego Arias It stands for "CERtified Approximate MembershIp STructures".
Nice :) and the pottery part? Is someone on the dev team a ceramist themselves?
Oh, that I don’t know. But I’ll try to ask Ilya :)
I just learned about this 8.11.1 breakage. I concur with Emilio that breaking a library in a patch-level release was certainly not intended. Was the source of the incompatibility clearly identified? From what I understood of the timeline, it was identified as soon as March 8th that there was an incompatibility with Coq master, so about one month before the 8.11.1 release and it would have been possible to prevent this issue to happen, if only Coq devs had known about this (apparently accidental) breakage.
Last updated: Feb 08 2023 at 07:02 UTC