- Feb 15, 2025
-
-
Anton Shestakov authored
-
- Mar 11, 2025
-
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
- Mar 08, 2025
-
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
- Mar 07, 2025
-
-
Anton Shestakov authored
-
- Feb 22, 2025
-
-
Anton Shestakov authored
-
- Nov 11, 2023
-
-
Anton Shestakov authored
Apparently it's required, even though the comment is kind of vague on this account? Anyway, the warning during docs building is gone.
-
- Feb 22, 2025
-
-
Anton Shestakov authored
-
- Mar 06, 2025
-
-
Anton Shestakov authored
-
- Jan 31, 2025
-
-
Pierre-Yves David authored
Upstream is getting a new parameter to deal with the lock hacks during "copy" clone. We need to adapt our wrapper. See mercurial-devel!1186
-
- Feb 19, 2025
-
-
Anton Shestakov authored
Python 3.13 was chosen simply as the newest possible version that our v2.1 CI images support. We install the build module right before actually building our packages to make sure it's always the latest version.
-
- Nov 16, 2024
-
-
Anton Shestakov authored
This should avoid incompatibility with other extension altering the dirstate. There are 2 cases here when we patch dirstate class on the fly: in repo._makedirstate() and in peer.local(). The first one is the "normal" way of making a dirstate for a repo and covers almost all cases, and the second one is special. In some cases during clone our dirstate is not correctly handled, so we need this band-aid solution just for clones on Mercurial 6.9 and older. See also 0f2268783c11 in core.
-
- Jan 31, 2025
-
-
Pierre-Yves David authored
Upstream is getting a new parameter to deal with the lock hacks during "copy" clone. We need to adapt our wrapper. See mercurial-devel!1186
-
- Feb 07, 2025
-
-
Anton Shestakov authored
-
Anton Shestakov authored
-
- Feb 06, 2025
-
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
Anton Shestakov authored
-
- Feb 04, 2025
-
-
Anton Shestakov authored
-
- Feb 03, 2025
-
-
Anton Shestakov authored
-
Anton Shestakov authored
-
- Feb 02, 2025
-
-
Anton Shestakov authored
-
Anton Shestakov authored
-
- Feb 01, 2025
-
-
Pierre-Yves David authored
If ctx is an instance of memctx, then ctx.node() returns None. Instead of trying to use node hash for diffing, we now use the ctx directly. Previously, this code was doing a diff involving the current working copy parent and gave the right output in the tests by total coincidence (since obslog is usually run with the working copy as the target revision). This is a two part fix as the previous changeset fixes the direction of the rebase that was wrong (an issue hidden by the bug fixed here).
-
Pierre-Yves David authored
We got the order of parents wrong, but this was hidden by another bug fixed in the next changeset. The idea here is that we rebase candidate from one of its parents onto a parent of node, so inside _rebase_cand() the logic is still correct: e.g. `candp1` is `old_base` and `nodep1` is `new_base`.
-
Pierre-Yves David authored
See f071b18e1382 and a502f3f389b5 in core.
-