- May 12, 2023
-
-
Georges Racinet authored
Following the change in GitLab Shell 14.13.0 (GitLab 15.6). This is explicitly not covered by tests.
-
- Apr 13, 2023
-
-
Georges Racinet authored
-
Georges Racinet authored
-
Georges Racinet authored
-
- Apr 12, 2023
-
-
Georges Racinet authored
The internal API now requires a JWT token derived from the shared secret, and for now also accepts only `gitlab-shell` as issuer. Finally, installation of the distribution in CI over the dusty `octobus/ci-base` image requires `ffi.h` (for `cryptography`, probably bindings to Rust code). It is simpler to use `python:3.9` instead, which installs `cffi` with a prebuilt wheel (and is based on Debian 11).
-
- Mar 31, 2023
-
-
Georges Racinet authored
We needed to push dev0 so that dependencies can be resolved in CI.
-
Georges Racinet authored
-
Georges Racinet authored
Now that we support the current stable versions of Mercurial and Evolve, we can do it.
-
Georges Racinet authored
For now, we need to refuse (non-default) topic namespaces, because we need a one-to-one mapping between branchmap entries and GitLab branches, and we have no way to represent the general case safely (partly because of conversion to Git, which is still performed in the vast majority of cases and prohibits the use of `//`).
-
Georges Racinet authored
Granted, this is trivial, but with the introduction of the "fully qualified branch name" in hg-evolve 11, this split will precisely become incorrect.
-
Georges Racinet authored
The only issue was the rename of `_filter_for_bookmarks` in mercurial/hg-git@f16d5105971d. It was done to avoid converting secret changesets, which are of no concern to us, since they won't be pushed to Heptapod by definition. Since we don't need the logic with the branch suffix, we're simply retrieving the hashes of all bookmarked changesets (which is what the `NoGitStateMaintainter` was already doing)
-
Georges Racinet authored
Way more convenient when it's a few statements than rerunning with html report and spining up a browser.
-
Georges Racinet authored
-
- Feb 13, 2023
-
-
Georges Racinet authored
This will simply have to be backed out when we switch over to *requiring* `hg-evolve>=11.0`
-
- Sep 05, 2022
-
-
Georges Racinet authored
-
Georges Racinet authored
The upgrade of the base CI image to flake8 5.0.4 apparently set the max line width to 79 chars (would have thought it to have come sooner). Only one offender so that's an easy fix
-
- Jul 08, 2022
-
-
Georges Racinet authored
In this case, the only change in the default branch was for compatibility with Mercurial 6.2 (not yet in use, but already tested in CI as head of stable branch)
-
- Jun 20, 2022
-
-
Georges Racinet authored
various CI improvements come with this merge, as well as a patch version bump for next release.
-
Georges Racinet authored
Compat tests: fixing test failing on current Mercurial stable See merge request !79
-
Georges Racinet authored
This failure affected the test itself only: `pycompat.bytesio` has been removed from Mercurial stable branch. We haven't been needing it since Heptapod became Python 3 only. This is a good example why Mercurial compat primitives should not be used in py-heptapod and hgitaly code bases.
-
Georges Racinet authored
The requirements file shipping with Heptapod is actually accepting patch version bumps for Mercurial and Evolve. Therefore we must test against their respecive heads of stable branches. (motivated by an actual test failure due to a compat change)
-
Georges Racinet authored
-
Georges Racinet authored
-
- Jun 12, 2022
-
-
Georges Racinet authored
-
Georges Racinet authored
-
- May 31, 2022
-
-
Georges Racinet authored
-
- May 16, 2022
-
-
-
Georges Racinet authored
At least with 6.2.5, pytest enforces that expected exceptions are subclasses of `BaseException`, which `error.Hint` is not. Thankfully we are now using Mercurial 6.1, which allows us to get back to the precise exception class without any version conditional.
-
- Apr 05, 2022
-
-
Georges Racinet authored
-
Georges Racinet authored
The publication of the final 1.0.0 version of hg-git brings in a minor incompatibility, that we will address once the current default branch becomes the stable branch (heptapod#658). This is not a problem for shipped versions because the requirements specified with the Rails app (both in heptapod and in release branch) already pin hg-git to 0.10.x
-
- Mar 14, 2022
-
-
Georges Racinet authored
-
Georges Racinet authored
-
- Feb 05, 2022
-
-
Georges Racinet authored
With the Mercurial 6.1 (core see heptapod#633) taking obsolescence into account while updating branchmap, it is no longer necessary to perform the (very slow) post-processing provided by the evolve extension. Because of the big jump in minimum required Mercurial version, we make it a major version change. Fixing heptapod#632 will thus boil down to requiring py-heptapod>=4.0.0dev0 This post-processing was the standard if evolve is activated (through the single heads enforcement logic), but the heptapod extension was (and still is) disabling it because it needs to call it itself anyway. See 834eb9379d73 for details.
-
- Feb 26, 2022
-
-
Georges Racinet authored
The new hgrc file creates some duplication with `required.hgrc` but at this point, an inclusion would probably be not worth the packaging risk.
-
Georges Racinet authored
We want the tests to pass on the version of hg-evolve that will eventually be shipped in the Heptapod release. Perhaps this indicates that we should have a permanent requirements file for dependencies to update on CI, but this will be for another day.
-
Georges Racinet authored
-
- Jan 25, 2022
-
-
Georges Racinet authored
[skip ci]
-
- Jan 23, 2022
-
-
Georges Racinet authored
-
- Jan 03, 2022
-
-
Georges Racinet authored
-
Georges Racinet authored
-