Commits
| Author | Commit | Message | Date | Builds |
|---|---|---|---|---|
Pierre-Yves David
|
commit-sorting: add sha as a tie breaker
In test, changeset will likely have the very same date. We need something to
stabilize the commit order.
|
|
||
|
clone/pull: filter refs more completely, also get tags
|
|
||
|
|
gitlog: most of a lazymanifest type
This lets us avoid a bunch of silly construction of full manifests in
cases where that's not needed.
|
|
||
|
|
506e927
M
|
Merge.
|
|
|
|
|
push: implement basic working push
|
|
||
|
|
clone: rework tests slightly and clean up clone implementation
We actually get all the refs now (I believe), and we push/pull to/from
a bare git repo in the test so we can actually push reliably.
|
|
||
|
|
cmdwrappers: add pull wrapper
|
|
||
|
|
foldme
|
|
||
|
|
cmdwrappers: move clone and init wrappers to new file
We're about to wrap even more commands, so let's collect them in a
single file.
|
|
||
|
|
clone: rewrite clone to actually work
The old version was very broken, and this is simpler and seems to work
well-enough.
|
|
||
|
|
gitlog: don't cache all manfiests/filedeltas
Hopefully this will help RAM use not go crazy.
|
|
||
|
Pierre-Yves David
|
log: enforce 'follow' when any file is provided
We cannot support bare log on file for now.
|
|
||
|
Pierre-Yves David
|
log/annotate: get something working
This is slow, and log only work with -f but this work.
|
|
||
|
Pierre-Yves David
|
gitdirstate: truncate float mtime to int
The dirstate/index content is truncated to int. So using the float version made
the dirstate write time appear newer than committed file.
|
|
||
|
|
bookwrap: stage writes and then commit them during bmstore.write
|
|
||
|
|
bookwrap: fix writing current mark
|
|
||
|
|
Drop obsolete bookmarks stuff.
|
|
||
|
|
Fix for hg bookmark output change.
|
|
||
|
|
tags: expunge tip tag
We don't like it in hg, and we don't like it here.
|
|
||
|
|
Fix up so that almost all tests pass against hg 3.4 and git 2.3.5.
|
|
||
|
|
gitlog: clean up refs iteration and blacklist refs/stash
|
|
||
|
|
gitlog: always use a set() of heads, instead of a sorted list
|
|
||
|
|
gitlog: actually update the cachedheads entry when done loading the cache
|
|
||
|
|
gitlog: tweaks to the phony manifest class to fix 404s in the web interface
I'm not sure why we end up in the manifest code here, but I'm not
terribly interested in chasing that down at the moment.
|
|
||
|
|
gitlog: less incorrect implementation of heads() on changelog
We still don't handle walking the graph forward from the specified
start point. Nothing I've noticed is broken yet.
|
|
||
|
|
gitlog: have rev() handle full hexlified shas
'hg heads -t' causes rev() to get a hexlified sha.
|
|
||
|
|
gitlog: sort commits more correctly
The old algorithm was just wrong: commits didn't end up topologically
sorted in the phony revlog, and the algorithm used to get interesting
commits near the tip of the repo didn't produce sensible results. This
uses a dumber heuristic which (at least on my local copy of dulwich)
produces more useful results.
The new algorithm should be able to handle incremental reindexing
slightly more gracefull…
|
|
||
|
|
Fix up most tests.
Only 'hg checkout' seems to be broken now.
|
|
||
|
|
gitrepo: fix up pathauditor and branchcache attrs on localrepo
|
|
||
|
|
hgit: make the testsuite at least not dump any tracebacks
Tests are still pretty heavily broken.
|
|