Skip to content
Snippets Groups Projects

update build system and README

Merged Ralf Jung requested to merge ralf/buildsys into master

We should have a release on opam ASAP so that we can point to it in the README. We could also have a "dev" version on opam for people that want to depend on that one, but not use our crazy opam.pins hack.

Merge request reports

Approval is optional

Merged by avatar (Apr 18, 2025 9:31am UTC)

Merge details

  • Changes merged into master with aad23129.
  • Deleted the source branch.

Pipeline #3837 passed

Pipeline passed for aad23129 on master

Activity

Filter activity
  • Approvals
  • Assignees & reviewers
  • Comments (from bots)
  • Comments (from users)
  • Commits & branches
  • Edits
  • Labels
  • Lock status
  • Mentions
  • Merge request status
  • Tracking
  • What needs to remain to be done to have a release on opam? I guess we should tag a stdpp1.0 and base the opam release on that?

  • Author Owner

    What remains to be done is essentially picking a versioning and branching scheme, and then picking a commit.

    The most widespread versioning scheme today is semantic versioning, where version numbers are triples x.y.z (major.minor.patch). The patch is incremented for bug-fix-only releases that are ideally forwards- and backwards-compatible, i.e., except for bugfixes, anything working with x.y.z1 also works with x.y.z2 and vice versa. Minor is upgraded for new features, so Anything that works with x.y keeps working with larger y. x is incremented for breaking changes. We don't have a real story for backwards-compatibility in Coq, but I still roughly use this scheme for Iris.

    The corresponding branching scheme then is that when version 1.0.0 is released, a branch 1.0 is created from which future 1.0.x releases are tagged.

    I personally prefer to have a separator between the version number and the name, e.g. stdpp-1.0.0 for tags and stdpp-1.0 or stdpp-1.0.z for branches, but that's obviously a matter of taste ;)

  • stdpp-x.y.z works for me.

    Another question is what will be our story on Coq 8.5 support. As shown in eda093d3, only very small changes are needed to make it work with Coq 8.5pl3. So, in principle, we could make sure that version 1.0.x works with both 8.5pl3 and 8.6.

  • Author Owner

    I'll leave that up to you. If you want to support Coq 8.5, I will happily adapt the CI to test for that.

    Edited by Ralf Jung
  • Alright, I have pushed 10c2f692 that makes it compile with both 8.5pl3 and 8.6. If you could setup the CI to test for both 8.5pl3 and 8.6 that would be great.

    Once we have tagged stdpp-1.0.0 and created a 1.0 branch I will revert 10c2f692 in master, so we can start using new Coq 8.6 features there.

  • Ralf Jung added 4 commits

    added 4 commits

    Compare with previous version

  • Author Owner

    Once we have tagged stdpp-1.0.0 and created a 1.0 branch

    1.0 branch or stdpp-1.0?

    If you could setup the CI to test for both 8.5pl3 and 8.6 that would be great.

    Done, I updated this PR and it is currently being tested on the CI.

  • Once we have tagged stdpp-1.0.0 and created a 1.0 branch 1.0 branch or stdpp-1.0?

    I was not very precise there, but let's follow the Iris naming conventions, i.e. call the branch stdpp-1.0 and the tag stdpp-1.0.0.

    Done, I updated this PR and it is currently being tested on the CI.

    Awesome!

  • Ralf Jung added 1 commit

    added 1 commit

    • 0274560f - update build system, CI and README

    Compare with previous version

  • Author Owner

    All right, this is all working fine and ready to be merged.

    Do you want "make validate" to be run on every CI run, only sometimes (e.g. randomly 10% of the cases), or never?

  • For Iris it's random, right? If so, let's do that here too.

  • Ralf Jung added 1 commit

    added 1 commit

    • 2d8a4eb4 - update build system, CI and README

    Compare with previous version

  • Author Owner

    done

  • mentioned in commit aad23129

  • Great, merged! I will create an issue for creating an OPAM package.

  • mentioned in issue #1 (closed)

Please register or sign in to reply
Loading