This change is replacing most of state.rules uses with state.actions
uses. The next change will change histeditstate class to actually
uses state actions.
the format of rules that we store in state file format is different from the rule
format that we present to users. We need a way of dumping action to state file.
To make histedit action objects responsible for understanding
the format of their action lines we are adding a torule method
which for a histedit action will return a string which can be
saved in histedit state or shown in text editor when editing the
plan.
This commits splits the parsing of the histedit rule from its semantic analysis.
It's necessary because sometimes we want to do first without doing the former (reading the
histedit state).
This decorator will is allowing us to move the registering the action
in actiontable closer to the action code. Also it is storing the
verb inside histedit action so the action is aware of the verb
needed to trigger it.
I want to refactor histedit to use action objects instead of (verb, rest)
pairs whenever possible. At the end of this series I want the rules to
be translated into action objects when reading state and translated back
when writing state. All histedit internals should use action objects instead
of state rules.
To migrate histedti internals sequentially I'm introducing the state.actions
property to translate rules on the fly so we can use both state.actions and
state.rules until refactoring is done.
Users may spend a lot of effort writing histedit rules,
getting an abort without being told they can recover their work
is very frustrating.
Avoid that by telling them where to find their work.
Used a class as a namespace, and then wired up a classmethod to return
all known constraints. I'm mostly happy with this, even though it's
kind of weird for hg.
This is a first (very simple) version of the histedit base action.
It works well in common usecases like rebasing the whole stack and
spliting the stack.
I don't see any obvious edge cases - but probably there is more than one.
That's why I want to keep it behind experimental.histeditng config knob
for now. I think on knob for all new histedit behaviors is better because
we will test all of them together and testers will need to turn it on only
once to get all new nice things.
For the future 'base' action in histedit we need a verification
constraint which will not allow using this action with changes
that are currently edited.
Before we can add a 'base' action to histedit need to change verification
so that action can specify which steps of verification should run for it.
Also it's everything we need for the exec and stop actions implementation.
I thought about baking verification into each histedit action (so each
of them is responsible for verifying its constraints) but it felt wrong
because:
- every action would need to know its context (eg. the list of all other
actions)
- a lot of duplicated work will be added - each action will iterate through
all others
- the steps of the verification would need to be extracted and named anyway
in order to be reused
The verifyrules function grows too big now. I plan to refator it in one of
the next series.
Before this patch we were using the old api bookmarks.write, this patches
replaces its usage by bookmarks.recordchange, the new api to record bookmark
changes.
Back in June we made histedit use obsolete markers to cleanup when possible.
This was rolled back as part of bb3db0db4037 (which should have only rolled back
the --abort stuff, but rolled back everything). This caused a nasty bug when
used in conjuction with the inhibit+directaccess extensions where histedit would
leave old nodes around even after they had been squashed away.
The root of the problem is that we first clean up old nodes, and then we clean
up temp nodes. In the first pass, when we obsoleted old nodes, some would become
unobsolete because they had temp nodes on top of them, thus making them stick
around even after the histedit finished.
The fix is to A) move the temp node cleanup to be before the old node cleanup
(since they are topological on top of the old nodes), and B) use obsolete
markers instead of stripping.
The home of 'Abort' is 'error' not 'util' however, a lot of code seems to be
confused about that and gives all the credit to 'util' instead of the
hardworking 'error'. In a spirit of equity, we break the cycle of injustice and
give back to 'error' the respect it deserves. And screw that 'util' poser.
For great justice.
When an user aborts a histedit, many things could go wrong. At a minimum, after
a histedit abort failure, their repository should be out of that state. We've
found situations where the user could not exit the histedit state without
manually deleting the histedit state file. This patch ensures that if any
exception happens during an abort, the histedit statefile will be deleted so
that users are out of the histedit state and can at least manually get the repo
back to a workable condition.
When the histeditstate class instance has it's clear() method called, there is
nothing to check to see if the state file exists before deleting it. It may not
exist, which would create an exception. This patch allows clear to be called at
any time.
This will be needed for the following patch.
If a histedit is progress, the 'histedit-state' file should exist. The patch
implements a convenience function to do check if a histedit is in progress.
This method will be use in next patch in the series.
This was the first ever feature request for histedit, originally filed
back on April 4, 2009. Finally fixed.
In the future we'll probably want to make it possible for other
preprocessing steps to be added to the list, but for now we're
skipping that because it's unclear what the API should look like
without a proposed consumer.
There is case where nodes are neither in tmpnodes nor leaf but still get
removed. For example, if you used the "edit" action, made a commit and run
--abort. The commit you made is not tracked by histedit, yet it will likely be
cleaned up with its parent. The commit may not tracked because no replacements
computations are done in the --abort case.
The for loop is already quite more complicated than necessary and we are about
to add some logic. Instead, we use a simple revset. Revset laziness should
provide us with similar performance.
The goal of this function is to strip content out of the repository. We do not
really care if this content is visible or cleanup node not and we should proceed
anyway. None of the internal actions are subject to this, however, a third party
extension running arbitrary commands during histedit is affected by this.
The process replacement is building a full mapping to allow moving bookmarks and
creating obsolescence marker. We do not need the full logic for abort so we
extract it. It will be useful as abort is missing some data about the
replacement and can crash when third party extensions push it a bit too far.
The faulty changeset use obsolescence marker to roll the repository back on
--abort. This is a problematic approach because --abort should be as close as an
actually transaction rollback as possible stripping all created data from the
repository (cf `hg rebase --abort` stripping all created changesets). Instead
3e883e7ec57b made all content created during the aborted histedit still
available in the repository adding obsolescence marker to make them hidden. This
will cause trouble to evolution user as a re-run of the same histedit (with
success) will likely result in the very same node to be "recreated" while
obsolescence marker would be in place for them. And canceling an obsoletion is
still a fairly complicated process.
This also rollback using obsmarkers instead of strip to clean up temporary node
on successful histedit run because the two change were not split in separated
changeset. Rolling that part back does not have significant consequence a will
have to be resubmitted independently
Before this patch, we were stripping temporary commits at the end of a histedit
whether it was successful or not. If we can create obs markers, we should
create them instead of stripping because it is faster and safer.
We use a variable to store whether or not we can create obsolescence markers.
It makes the patch series more readable as we are going to reuse this
values in other places in the function.
Python 2.6 introduced the "except type as instance" syntax, replacing
the "except type, instance" syntax that came before. Python 3 dropped
support for the latter syntax. Since we no longer support Python 2.4 or
2.5, we have no need to continue supporting the "except type, instance".
This patch mass rewrites the exception syntax to be Python 2.6+ and
Python 3 compatible.
This patch was produced by running `2to3 -f except -w -n .`.
This is suboptimal as the user still has to explicitly cancel the
histedit afterwards, but it prevents the immediate problem.
histedit should probably implicitly do 'hg histedit --abort' if a
util.Abort is raised internally.
The phrase "cannot edit immutable changeset" is kind of tautological.
Of course unchangeable things can't be changed. We instead mention
"public" and provide a hint so that we can point to the actual
problem. Even in cases where some operation other than edition cannot
be performed, "public" gives the root cause that results in the
"immutable" effect.
There is a precedent for saying "public" instead of "immutable", for
example, in `hg commit --amend`.
If histedit failed after all the rules were complete (for instance, if there was
an exception in the cleanup phase), you couldn't --continue because it was
unable to pop a rule. Let's just skip the rule execution phase of --continue if
there are no more rules.
If the histedit backupfile was None (like if evolve is enabled) it would get
serialized as 'None' into the state file. Later if the histedit was aborted and
the top most commit was unreachable (ex: if it was obsolete or stripped),
histedit would try to unbundle the backupfile and try to read .hg/None.
This fixes it to not serialize None. Since it only happens with evolve, I'm not
sure how to add test coverage here.
--edit-plan was completely broken from the command line because it used an old
api that was not updated (it would crash with a stack trace). Let's update it
and add tests to catch this.
Commit 960f8ca79ab1 broke histedit's rollup by causing it to open the editor.
Turns out I missed a spot where the rollup option was read.
This fixes that and adjusts the test to catch this case.
The existing state serialization format assumed the rule line consisted of an
action and a hash. In our external extension that adds 'exec' this is not the
case (there is no hash, just the shell command). So let's change the format to
be more generic with just an action and a remainder, and the various commands
can handle it as they wish.
Flagged for stable since we want to get this format tweak in before the new
format goes live in the release.
Previously the fold action would inspect it's class to figure out if it was a
rollup or not. This was hacky. Now that finishfold is inside the fold class,
let's modify it to check a function (which roll can override) to determine if it
should be prompting for a commit message.
This converts the fold/roll actions into a histeditclass instance, as part of an
ongoing effort to refactor histedit for maintainability and robustness.
The tests changed for two reasons:
1) We get a new 'empty changeset' warning because we now warn more consistently
between normal histedit and --continue about commits disappearing.
2) Previously we were not putting the histedit-source extra field on the
temporary fold commit during normal runs, but we were on --continue runs. By
unifying these code paths we now consistently put histedit-source on the
temporary fold commit, which changes some of the hashes in the backup bundles.
This converts the pick action into a histeditclass instance, as part of an
ongoing effort to refactor histedit for maintainability and robustness.
The test output changed because previously pick would only report the commit
disappearing if there were no merge conflicts. Now that we've unified the normal
and the --continue flows, they act the same and we get the warning even after
--continue.
This takes the newly added histeditaction class and integrates it into the
histedit run and bootstrapcontinue logic. No actions implement the class yet,
but they will be add in upcoming patches.
This adds a new class called histeditaction. It represents a single action in a
histedit plan. Future patches will integrate it into the histedit flow, then
convert each existing action in to the class form one by one.
This is part of a larger refactor aimed at increasing histedit robustness,
maintainability, and extensibility.
Previously we would not allow --continue if the current working copy parent was
not a descendant of the commit produced by the previous histedit step. There's
nothing really blocking us from continuing the histedit in this situation, so
let's stop aborting in this case.
This is technically a BC change, but it is making things more forgiving so I
think it's ok.
In the future we will likely add an 'exec' action to histedit, which means the
user can do whatever they want during the middle of a histedit (like perhaps
calling 'hg update'), which means we'll need to support this case eventually
anyway.
It's possible for the user to delete some of the commits they started with
during a histedit, and aborting the histedit doesn't bring them back. Let's
store a backup bundle so we can always recover the stack of commits from before
they began.
We were trying to prevent strips of important nodes during histedit,
but the check was actually comparing the short hashes in the rules to
the exact value the user typed in, so it only ever worked if the user
typed a 12 character hash.
Extension authors (notably at companies using hg) have been
cargo-culting the `testedwith = 'internal'` bit from hg's own
extensions, which then defeats our "file bugs over here" logic in
dispatch. Let's be more aggressive about trying to give extension
authors a hint about what testedwith should say.
In a previous commit we removed the extra histedit object instance being
constructed in --continue and --abort. The new --edit-todo missed this fix
though (which means the state object it produces doesn't have the locks on it).
It's not breaking anything now, but let's go ahead and clean that up before we
forget.
Before this change hg help histedit would use the default variable label:
--commands VALUE
...
-r --rev VALUE [+]
With this change the text will be in the usual help text style and a bit more
explanatory:
--commands FILE
...
-r --rev REV [+]
The rest of help messages for command arguments are simple phrases without any
ending punctuation, so having this text be a complete sentence didn't really
fit.
Since many users are using terminals wider than 80 chars there should be an
option to have longer lines in histedit editor.
Even if the summary line is shorter than 80 chars after adding action line
prefixes (like "pick 7c2fd3b9020c") it doesn't fit there anymore. Setting
it to for example 110 would be a nice option to have.
Previously, the histedit state object was being recreated during continue/abort.
This meant that the locks that were held on the original state object were not
available to actions, which meant actions could not release the lock on the
repository (like an 'exec' action would need to do).
This affected our internal extension that added the 'exec' action.
Currently, if the node no longer exists, the state object fails to load
and pukes with an exception. Changing the state object to only store the
node allows callers to handle these cases. For instance, in
bootstrapcontinue we can now detect that the node doesn't exist and exit
gracefully.
The alternative is to have the state object store something like None
when the node doesn't exist, but then outside callers won't be able to
access the old node for recovery (unless we store both the node and the
ctx, but why bother).
More importantly it allows us to detect this case when doing hg histedit
--abort. Currently this situation results in both --continue and
--abort being broken and the user has to rm .hg/histedit-state to unwedge
their repo.
(description by Durham Goode)
During histedit we don't want user to do any operation resulting in
stripping nodes needed to continue history editing. This patch
wraps the strip function to detect such situations.
Adds a configuration setting for allowing users to specify the default behavior
of 'hg histedit' without arguments. This saves users from having to manually
figure out the bottom commit or a complicated revset. My current revset of
choice is "only(.) & draft() - ::merge()"
The commits that histedit can work with is usually quite limited, so if this
feature ends up working well, we may want to consider making "only(.) & draft()
- ::merge()" the default behavior for everyone.
Previously histedit only stored the short version of the rule nodes in the
state. This meant that later we couldn't resolve a rule node to its full
form if the commit had been deleted from the repo.
Let's store the full form from the beginning.
Read the state in histeditstate. This allows us to correctly update
internal variables when necessary without having to recreate a new
state. When we read a state in _histedit state while we will already
have state passed from histedit(), we can read the state in place
and don't have to merge two histeditstates.
The histedit code often expects a context. However histedit hands
around the tuple for the serialization and therefore hand over a
parentctxnode. This leads to code having to return a context based
on the parentctxnode. We let the state only return a context but
correctly serialize and deserialze to a node.
Add an histeditstate class that is intended to hold the current
state. This allows us encapsulate the state and avoids passing
around a tuple which is based on the serialization format. In
particular this will give actions more control over the state and
allow external sources to have more control of histedits behavior,
e.g. an external implementation of x/exec.
The basic obsolete option is allowing the creation of obsolete markers. This
does not enable other features, such as allowing unstable commits or exchanging
obsolete markers.
The `nodes` object is a set. We sort it to get stable order. This is going to
prevent revsets from getting confused when removing a `.set()` call in `roots`.
This wraps all the locations of dirstate.setparent with the appropriate
begin/endparentchange calls. This will prevent exceptions during those calls
from causing incoherent dirstates (issue4353).
The internal API used IOError to indicate that a file should be marked as
removed.
There is some correlation between IOError (especially with ENOENT) and files
that should be removed, but using IOErrors to represent file removal internally
required some hacks.
Instead, use the value None to indicate that the file not is present.
Before, spurious IO errors could cause commits that silently removed files.
They will now be reported like all other IO errors so the root cause can be
fixed.
When the authorship of the changeset folded in does not match that of
the base changeset, we currently use the configured ui.username
instead. This is especially surprising when the user is not the author
of either of the changesets. In such cases, the resulting authorship
(the user's) is clearly incorrect. Even when the user is folding in a
patch they authored themselves, it's not clear whether they should
take over the authorship. Let's instead keep it simple and always
preserve the base changeset's authorship. This is also how
"git rebase -i" handles folding/squashing.
This new histedit command (short for "rollup") is a variant of "fold" akin to
"hg amend" for working copy: it accumulates changes without interrupting
the user and asking for an updated commit message.
This patch passes 'editform' argument according to the format below:
EXTENSION[.COMMAND][.ROUTE]
- EXTENSION: name of extension
- COMMAND: name of command, if there are two or more commands in EXTENSION
- ROUTE: name of route, if there are two or more routes for COMMAND
In this patch:
- 'edit', 'fold', 'mess' and 'pick' are used as COMMAND
- ROUTE is omitted
'histedit.pick' case is very rare, but possible if:
- target revision causes conflict at merging (= requires '--continue'), and
- description of it is empty ('hg commit -m " "' can create such one)
In the code path for 'histedit --continue' (the last patch hunk),
'canonaction' doesn't contain the entry for 'fold', because 'fold'
action causes:
- using temporary commit message forcibly, and
- making 'editopt' False always (= omit editor invocation if commit
message is specified)
Before this patch, trimming description of each changesets in histedit
may split at intermediate multi-byte sequence.
This patch uses 'util.ellipsis' to trim description of each changesets
instead of directly slicing byte sequence.
Even though 'util.ellipsis' adds '...' as ellipsis when specified
string is trimmed (= this changes result of trimming), this patch uses
it, because:
- it can be used without any additional 'import', and
- ellipsis seems to be better than just trimming, for usability
This patch changes the calling signature of memfilectx's __init__ to fall in
line with the other file contexts.
Calling code and tests have been updated accordingly.
This omits (redundant) adding "\n' to "message", because:
- empty line is inserted by "commitforceeditor", if editor is invoked
- tail white-spaces are stripped at storing into chaneglog, otherwise
This patch also enhances "test-histedit-edit.t", because "hg histedit"
hasn't been explicitly tested around editor invocation and
"--continue" option.
Before this patch, "hg histedit" for "message" uses "ui.edit()" for
commit message editing.
It shows original commit message, but not detail about the target
revision: status of each modified/added/removed files, for example.
This patch uses the editor gotten by "getcommiteditor()" instead of
"ui.edit()" for "message"
In "test-histedit-edit.t", this patch omits "fixbundle" invocation,
because it prevents from confirming the "HG: added f" line in commit
message by filtering " added " lines.
Omiting "fixbundle" invocation causes that the exit code of "hg
histedit" appears as one of command line: in this case, "hg histedit"
is aborted by (expected) exception raising.
The preceding patch causes that "memctx.__init__()" with "editor"
argument invokes editor and saves edited commit message into
".hg/last-message.txt".
This patch passes "editor" argument to "memctx.__init__()" instead of
explicit invocations of "commitforceeditor()" and
"savecommitmessage()" for "collapse" command.
Before this patch, "message" action of "hg histedit" uses "ui.edit()"
explicitly to get commit message edited manually.
This requires explicit "localrepository.savecommitmessage()"
invocation to save edited commit message into ".hg/last-message.txt",
because unexpected exception raising may abort command execution
before saving it in "localrepository.commit()".
This patch uses "editor" argument of "localrepository.commit()"
instead of explicit "ui.edit()" invocation for "message" action of "hg
histedit"
"localrepository.commit()" will invoke "editor()" function newly added
in this patch, and save edited commit message into ".hg/last-message.txt"
automatically.
Before this patch, manually edited commit message for "message"
command in histedit-ing is not saved into ".hg/last-message.txt" until
it is saved by "localrepository.savecommitmessage()" in
"localrepository.commit()".
This may lose such commit message, if unexpected exception is raised.
This patch saves manually edited commit message for "message" comand
in histedit-ing into ".hg/last-message.txt" just after user editing.
This is the simplest implementation to fix on stable. Editing and
saving commit message should be centralized into the framework of
"localrepository.commit()" with "editor" argument in the future.
This patch uses repository wrapping class for exception raising before
saving commit message in "localrepository.commit()" easily and
certainly, because such exception requires corner case condition.
Before this patch, manually edited commit message for "fold" command
in histedit-ing is never saved into ".hg/last-message.txt", because it
uses "localrepository.commitctx()" instead of
"localrepository.commit()": saving into ".hg/last-message.txt" is
executed only in the latter.
This patch saves manually edited commit message for "fold" command in
histedit-ing into ".hg/last-message.txt" just after user editing.
This is the simplest implementation to fix on stable. Editing and
saving commit message for memctx should be centralized into the
framework like "localrepository.commit()" with "editor" argument or so
in the future.
When we specify a revision or a revset we just get the last element from the
list. For revsets this can lead to unintended effects where you specify a
revset like only() but instead histedit selects the highest revision in the
set as root. Therefore we should always use the lowest revision number as
root.
Some extensions set configuration settings that showed up in 'hg showconfig
--debug' with 'none' as source. That was confusing.
Instead, they will now tell which extension they come from.
This change tries to be consistent and specify a source everywhere - also where
it perhaps is less relevant.
Currently, histedit acquires and releases lock and wlock several times during
its run. This isn't great because it allows other hg processes to come in and
change state. With this fix, lock and wlock are acquired and released exactly
once.
The change to test-histedit-drop.t is a minor implementation one -- the cache
is still correctly invalidated, but it just happens a little later and only
gets printed out because of the unrelated --debug flag.
Mercurial earlier than 2.7 allows users to do anything other than
starting new histedit, even though current histedit is not finished or
aborted yet. So, unfinished (and maybe inconsistent now) histedit
states may be left and forgotten in repositories.
Before this patch, histedit extension shows the message below, when it
detects such inconsistent state:
abort: REV is not an ancestor of working directory
(update to REV or descendant and run "hg histedit --continue" again)
But this message is incorrect, unless old Mercurial is re-installed,
because Mercurial 2.7 or later disallows users to update the working
directory to another revision.
This patch changes the hint message to suggest "hg histedit --abort".
Before this patch, if there are multiple roots in "--outgoing"
revisions, result of "histedit --outgoing" depends on the parent of
the working directory. It succeeds only when the parent of the working
directory is a descendant of the oldest root in "--outgoing"
revisions, and fails otherwise.
It seems to be ambiguous and difficult for users.
This patch makes "histedit --outgoing" abort if there are multiple
roots in "--outgoing" revisions always.
Before this patch, if there are multiple roots in "--outgoing"
revisions, result of "histedit --outgoing" depends on the parent of
the working directory. It succeeds only when the parent of the working
directory is a descendant of the oldest root in "--outgoing"
revisions, and fails otherwise.
It seems to be ambiguous and difficult for users.
This patch makes "histedit --outgoing" abort if there are multiple
roots in "--outgoing" revisions always.