Recently we switched rebases to run the entire rebase inside a single
transaction, which dramatically improved the speed of rebases in repos with
large working copies. Let's also move the dirstate into a single dirstateguard
to get the same benefits. This let's us avoid serializing the dirstate after
each commit.
In a large repo, rebasing 27 commits is sped up by about 20%.
I believe the test changes are because us touching the dirstate gave the
transaction something to actually rollback.
The way default marker template was defined before this patch,
the spacing before dash in conflict markes was dependent on
whether changeset is a tip one or not. This is a relevant part
of template:
'{ifeq(tags, "tip", "", "{tags} "}'
If revision is a tip revision with no other tags, this would
resolve to an empty string, but for revisions which are not tip
and don't have any other tags, this would resolve to a single
space string. In the end this causes weirdnesses like the ones
you can see in the affected tests.
This is a not a big deal, but double spacing may be visually
less pleasant.
Please note that test changes where commit hashes change are
the result of marking files as resolved without removing markers.
Before this patch collapse message wasn't stored so when
you ran into the merge conflict while rebasing, running
rebase --continue didn't remember the message and
always opened editor to fill commit message.
This patch adds saving collapse message in
.hg/last-message.txt and restoring it later
when needed.
As we will make 'rebase' behave more like 'merge', it will no longer pick
'max(branch(.))' as the default destination. We have to hard code the expected
destination is multiple tests where it matters. After a careful inspection none
of theses tests really cares about the default destination behavior and just
omitted one out of laziness.
--collapse will do that rebase doesn't commit until the final commit. The lack
of a new commit would make it look like the rebase didn't contribute any
changes.
Instead, only warn about no commits when not using --collapse.
This warning exists to prevent git users from prematurely polluting
their namespace when trying out Mercurial. But for repos that already
have multiple branches, understanding what branches are is not
optional so we should just shut up.
Previously, a backup bundle could overwrite an existing bundle and cause user
data loss. For instance, if you have A<-B<-C and strip B, it produces backup
bundle B-backup.hg. If you then hg pull -r B B-backup.hg and strip it again, it
overwrites the existing B-backup.hg and C is lost.
The fix is to add a hash of all the nodes inside that bundle to the filename.
Fixed up existing tests and added a new test in test-strip.t
Show status messages while rebasing, similar to what graft do:
rebasing 12:2647734878ef "fork" (tip)
This gives more context for the user when resolving conflicts.
Globbing the hash made it harder to maintain tests with run-tests -i when it
was so far by the generated test output.
The hashes are stable and we just need to add a (glob).
Before this patch, if both "--message" and "--collapse" are specified
for "hg rebase", "rebaes.normal" is used as "editform" unexpectedly.
Unlike patches before and after in this series for improvement, this
is bug fix patch.
Before this patch, "rebase --collapse --edit" without "--message" and
"--logfile" invokes editor twice unexpectedly:
1. explicit "ui.edit()" invocation in rebase extension itself
2. indirect invocation in "localrepository.commit()" with "editor =
commitforceeditor" assigned by "--edit" option
This patch uses indirect "commitforceeditor" invocation instead of
"ui.edit()" for "--collapse" without "--message" and "--logfile" to:
- suppress redundant the former invocation
- ensure editor invocation even when "--edit" is not specified
Rebasing with --collapse would leave the working copy on the parent of the
collapsed commit, instead of on the collapsed commit. This fixes that. Also
fixes a few tests that already covered this area but had bad data.
This also fixes issue3716 where bookmarks are not kept across rebases with
--collapse. I updated the test to cover that case as well.
Prior to this changeset, rebase always left the working directory as a parent of
the last rebased changeset. The is dubious when, before the rebase, the working
directory was not a parent of the tip most rebased changeset.
With this changeset, we move the working directory back to its original parent.
If the original parent was rebased, we use it's successors.
This is a step toward solving issue3813 (rebase loses active bookmark if it's
not on a head)
In the way to solving issue3813 we'll preserve the working directory parent
after the rebase. Multiple rebases test expect the working directory parent to
be tip after rebase. We patches them before the actual change to prevent
confusion.
When a series of commits first adds a file and then removes it,
hg rebase --collapse prompts whether to keep the file or delete it. This is
due to it reusing the branch merge code. In a noninteractive terminal it
defaults to keeping the file, which results in a collapsed commit that is
has a file that should be deleted. This bug resulted in developers accidentally
commiting unintentional changes to our repo twice today, so it's fairly
important to get fixed.
This change allows rebase --collapse to tell the merge code to accept the
latest version every time without prompting.
Adds a test as well.
'export' is the official export format and used by patchbomb, but it would only
show date as a timestamp that most humans might find it hard to relate to. It
would be very convenient when reviewing a patch to be able to see what
timestamp the patch will end up with.
Mercurial has always used util.parsedate for parsing these headers. It can
handle 'all' date formats, so we could just as well use a readable one.
'export' will now use the format used by 'log' - which is the format described
as 'Unix date format' in the templating help. We assume that all parsers of '#
HG changeset patch'es can handle that.
Bookmarks/branches/tags shouldn't be allowed to be integers because that
overlaps with revision numbers. Right now if a user created one they can't
use it anyway because the revision numbers take precedence.
The check only happens when creating a new bookmark/etc from a command so it
shouldn't affect existing bookmarks/branches/tags or importing branches from
git.
This fix was prompted by us having a user create a bookmark named "404" then
accidentally checkout a very old version of our repository.
The strip code used a trick to lower the cost of branchcache update after a
strip. However is less necessary since we have branchcache collaboration.
Invalid branchcache are likely to be cheaply rebuilt again a near subset of the
repo.
Moreover, this trick would need update to be relevant in the now filtered
repository world. It currently update the unfiltered branchcache that few people
cares about. Make it smarter on that aspect would need complexes update of the
calling logic
So this mechanism is:
- Arguably needed,
- Currently irrelevant,
- Hard to update
and I'm dropping it.
We now update the branchcache in all case by courtesy of the read only reader.
This changeset have a few expected impact on the testsuite are different cache
are updated.
Now that changelog filtering is in place, it's become evident that
naming the filters according to the set of revs _not_ included in the
filtered changelog is confusing. This is especially evident in the
collaborative branch cache scheme.
This changes the names of the filters to reflect the revs that _are_
included:
hidden -> visible
unserved -> served
mutable -> immutable
impactable -> base
repoview.filteredrevs is renamed to filterrevs, so that callers read a
bit more sensibly, e.g.:
filterrevs('visible') # filter revs according to what's visible
Filtered repository are *subset* of unfiltered repository. This means that a
filtered branchmap could be use to compute the unfiltered version.
And filtered version happen to be subset of each other:
- "all() - unserved()" is a subset of "all() - hidden()"
- "all() - hidden()" is a subset of "all()"
This means that branchmap with "unfiltered" filter can be used as a base for
"hidden" branchmap that itself could be used as a base for unfiltered
branchmap.
unserved < hidden < None
This changeset implements this mechanism. If the on disk branchcache is not valid
we use the branchcache of the nearest subset as base instead of computing it from
scratch. Such fallback can be cascaded multiple time is necessary.
Note that both "hidden" and "unserved" set are a bit volatile. We will add more
stable filtering in next changesets.
This changeset enables collaboration between no filtering and "unserved"
filtering. Fixing performance regression introduced by 7bff5f37cb97
This function augments strip to incrementally update the branchheads cache
rather than recompute it from scratch. This speeds up the performance of strip
and rebase on repos with long history. The performance optimization only
happens if the revisions stripped are all on the same branch and the parents of
the stripped revisions are also on that same branch.
This adds a few test cases, particularly one that reproduces the extra heads
that mpm observed.
Rebase now behaves as if --detach was always passed. Non-merges are
rebased as non-merges, regardless of their parent being an ancestor of
the destination. Merges will usually be rebased as merges unless both of
their parents are ancestors of the destination, or one of their parents
is pruned when rebased.
This only alters the behavior of rebase when using the --source/--rev
options. --detach option is deprecated.
All test changes were carefully validated.
We allow rebase plus collapse, but not collapse only? I imagine people would
rebase first then collapse once they are sure the rebase is correct and it is
the right time to finish it.
I was reluctant to submit this patch for reasons detailed below, but it
improves rebase --collapse usefulness so much it is worth the ugliness.
The fix is ugly because we should be fixing the collapse code path rather than
the merge. Collapsing by merging changesets repeatedly is inefficient compared
to what commit --amend does: commitctx(), update, strip. The problem with the
latter is, to generate the synthetic changeset, copy records are gathered with
copies.pathcopies(). copies.pathcopies() is still implemented with merging in
mind and discards information like file replaced by the copy of another,
criss-cross copies and so forth. I believe this information should not be lost,
even if we decide not to interpret it fully later, at merge time.
The second issue with improving rebase --collapse is the option should not be
there to begin with. Rebasing and collapsing are orthogonal and a dedicated
command would probably enable a better, simpler ui. We should avoid advertizing
rebase --collapse, but with this fix it becomes the best shipped solution to
collapse changesets.
And for the record, available techniques are:
- revert + commit + strip: lose copies
- mq/qfold: repeated patching() (mostly correct, fragile)
- rebase: repeated merges (mostly correct, fragile)
- collapse: revert + tag rewriting wizardry, lose copies
- histedit: repeated patching() (mostly correct, fragile)
- amend: copies.pathcopies() + commitctx() + update + strip
The fix introduced in 3509b9cf8f86 was only partially successful. It is correct
to turn dirstate 'm' merge records into normal/dirty ones but copy records are
lost in the process. To adjust them as well, we need to look in the first
parent manifest to know which files were added and preserve only related
records. But the dirstate does not have access to changesets, the logic has to
moved at another level, in localrepo.
The buildstate function now take a set of revs. Logic related to --source and
--base option have been moved in the main rebase function.
In the process this fixes a bug where the wrong source changeset might be pick.
This explain the changes in hgext/rebase.py
Add two changesets to the scenario so that the bundle can be reused
within three tests.
Before:
@ 5: 'F'
|
| o 4: 'E'
|/|
o | 3: 'D
| |
| o 2: 'C'
|/
| o 1: 'B'
|/
o 0: 'A'
After:
@ 7: 'H'
|
| o 6: 'G'
|/|
o | 5: 'F'
| |
| o 4: 'E'
|/
| o 3: 'D'
| |
| o 2: 'C'
| |
| o 1: 'B'
|/
o 0: 'A'
Revisions 0-1 keep the same number/label. Others were translated by
an offset of 2 (2.C -> 4.E)
When collapsing changesets with rebase, you get a chance to edit the commit
message manually, but there is no way to pass this message from the command
line. This patch adds a `--message` (with short form `-m`) and `--logfile`
(with short form `-m`) options to the rebase command. These options suppresses
the generation of the default commit message, and instead use the message
provided in the option (in case of `-m`) or in the file it points to (in case
of `-l`).
If you use this option without the `--collapse` option, it will raise an
error.
Options documentation edited by Patrick Mezard <pmezard@gmail.com>
