coq-community / coq-nix-toolbox

Nix helper scripts to automate local builds and CI [maintainers=@CohenCyril,@Zimmi48]
MIT License
32 stars 9 forks source link

mathcomp: 1 -> 2 #241

Closed proux01 closed 3 weeks ago

proux01 commented 1 month ago

Diff is not very readable so here is a sumup of the change in packages that are available with default mathcomp

8.16
- coq-bits
- coqeal
- deriving
- extructures
- gaia
- gaia-hydras
- mathcomp-analysis
- mathcomp-apery
- mathcomp-classical
- mathcomp-infotheo
- multinomials

=> let's keep MC1 the default on Coq 8.16

8.17
+ mathcomp-tarjan

8.18
+ mathcomp-tarjan
+ ssprove

8.19
+ coqeal
+ gaia
+ graph-theory
+ mathcomp-tarjan
+ ssprove

8.20
+ coqeal
+ gaia
+ mathcomp-algebra-tactics
+ mathcomp-analysis
+ mathcomp-classical
+ mathcomp-real-closed
+ mathcomp-tarjan
CohenCyril commented 1 month ago

@proux01 Given the losses for 8.16, I would advocate keeping mathcomp 1 as default there. As for 8.17 it's on par, but out of sentimentality I would be inclined to keep mathcomp 1 so that gaia still compiles there. From 8.18 on I am in favor of switching to mathcomp 2.

CohenCyril commented 1 month ago

Thanks for making the experiments.

proux01 commented 1 month ago

Makes a lot of sense. About gaia, it appears that we just never updated the Nix package with the newer release on MathComp 2, then let's switch to MC2 starting at Coq 8.17 since it's a net gain from there.

CohenCyril commented 1 month ago

Makes a lot of sense. About gaia, it appears that we just never updated the Nix package with the newer release on MathComp 2, then let's switch to MC2 starting at Coq 8.17 since it's a net gain from there.

Ah perfect! Well done

proux01 commented 1 month ago

So CI seems happy and confirms that we don't remove any package