The previous commit already fixed the user-visible lag, but there's still a
problem with multiple background git processes consuming resources calculating
diffs that we are never going to show. Improve this by terminating those
processes (by sending them a TERM signal).
Unfortunately this is only possible on Linux and Mac, so Windows users will have
to live with the higher CPU usage. The recommended workaround is to not use
"diff.algorithm = histogram".
As we just did for tasks, close their stdout pipe instead. This makes the called
process terminate more gracefully.
This isn't a change that we *need* to make, it's just a bit nicer.
Previously we would call git merge-base with the upstream branch to determine
where unpushed commits end and pushed commits start, and also git merge-base
with the main branch(es) to see where the merged commits start. This worked ok
in normal cases, but it had two problems:
- when filtering by path or by author, those merge-base commits would usually
not be part of the commit list, so we would miss the point where we should
switch from unpushed to pushed, or from pushed to merged. The consequence was
that in filtering mode, all commit hashes were always yellow.
- when main was merged into a feature branch, we would color all commits from
that merge on down in green, even ones that are only part of the feature branch
but not main.
To fix these problems, we switch our approach to one where we call git rev-list
with the branch in question, with negative refspecs for the upstream branch and
the main branches, respectively; this gives us the complete picture of which
commits are pushed/unpushed/merged, so it also works in the cases described
above.
And funnily, even though intuitively it feels more expensive, it actually
performs better than the merge-base calls (for normal usage scenarios at least),
so the commit-loading part of refresh is faster now in general. We are talking
about differences like 300ms before, 140ms after, in some unscientific
measurements I took (depends a lot on repo sizes, branch length, etc.). An
exception are degenerate cases like feature branches with hundreds of thousands
of commits, which are slower now; but I don't think we need to worry about those
too much.
This makes it easier to use the full ref in the git merge-base call, which
avoids ambiguities when there's a tag with the same name as the current branch.
This fixes a hash coloring bug in the local commits panel when there's a tag
with the same name as the checked out branch; in this case all commit hashes
that should be yellow were painted as red.
GetMergeBase is always called with a full ref, so it shouldn't need the
ignoringWarnings hack (which is about ignoring warnings coming from ambiguous
refs).
Also, separate stdout and stderr, which would also have solved the problem. We
no longer really need it now, but it's still cleaner.
Also, fix two other commands that stage all files under the hood:
- when continuing a rebase after resolving conflicts, we auto-stage all files,
but in this case we never want to include untracked files, regardless of the
filter
- likewise, pressing ctrl-f to find a base commit for fixup stages all files for
convenience, but again, this should only stage files that are already tracked
It seems that `git for-each-ref` is a lot slower than `git tag --list` when
there are thousands of tags, so revert back to the previous method, now that we
no longer use the IsAnnotated field.
This reverts commit b12b1040c3.
Storing it in the Tag struct makes loading tags a lot slower when there is a
very large number of tags; so determine it on the fly instead. On my machine,
the additional call takes under 5ms, so it seems we can afford it.
Like message, extraField can get very long (when there are thousands of tags on
a single commit), so move it to the end; this allows us to truncate overly long
lines in the output and still get all the essential fields.
Co-authored-by: Stefan Haller <stefan@haller-berlin.de>
When entering staging (or patch building) for an added or deleted file, it
doesn't make sense to use hunk mode, because pressing space would stage/unstage
the entire file, and if the user wanted to do that, they would have pressed
space in the Files panel. So always use line mode for added/deleted files by
default, even if the useHunkModeInStagingView user config is on.
Not because it's terribly important to test here (doesn't hurt though), but
because it will be useful in the next commit for a new test we're adding there.
When filtering for a file path we use the --follow option for "git log", so it
will follow renames of the file, which is great. However, if you then selected
one of the commits before a rename, you didn't see a diff, because we would pass
the original filter path to the "git show" call.
To fix this, call git log with the --name-status option when filtering by path,
so that each commit reports which file paths are touched in this commit;
remember these in the commit object, so that we can pass them to the "git show"
call later.
Be careful not to store too many such paths unnecessarily. When filtering by
folder rather than file, all these paths will necessarily be inside the original
filter path, so detect this and don't store them in this case.
There is some unfortunate code duplication between loading commits and loading
reflog commits, which I am too lazy to clean up right now.
These are very similar in that they call RunAndProcessLines on a git log command
and then process lines to create commits. Extract this into a common helper
function. At the moment this doesn't gain us much, but in the next commit we
will extend this helper function considerably with filter path logic, which
would otherwise have to be duplicated in both places.
It is good practice to use a -- argument even if no further arguments follow.
Doesn't really make a difference for this particular command though, I think.
This broke with the introduction of the age in the stashes list in bc330b8ff3
(which was included in 0.41, so 12 minor versions ago).
This makes it work again when filtering by file, but it still doesn't work when
filtering by folder (and never has). We'll fix that next.
This was not a good idea, and it seems it has never been tested: the --name-only
and -z options don't work well together. Specifically, -z seems to simply cancel
--name-only.
This is not enough to make it work, we'll need more fixes in the next commit.
This reverts commit 50044dd5e0.
At the same time, we change the defaults for both of them to "date" (they were
"recency" and "alphabetical", respectively, before). This is the reason we need
to touch so many integration tests. For some of them I decided to adapt the test
assertions to the changed sort order; for others, I added a SetupConfig step to
set the order back to "recency" so that I don't have to change what the test
does (e.g. how many SelectNextItem() calls are needed to get to a certain
branch).
When toggling the value in the UI we simply overwrite the value in UserConfig;
this would be bad if there was ever a chance that we want to write the user
config back to disk, but it is very unlikely that we can do that, because
currently we have no way to tell which parts of the config come from the global
config file and which ones come from a repo-local one.
While it's true that the behavior is a little different from the staging panel,
where the staged lines are actually removed from the view and in many cases the
selection stays more or less in the same place, it is still very useful to move
to the next stageable thing in the custom patch building view too.
This improves the experience when selecting a hunk generously with the mouse, by
dragging over it including some context lines above and below. Previously we
would consider the "moving end" of the selection range for whether things need
to be added or removed, but this doesn't make sense if it's a context line. Now
we consider the first actual change line that is included in the range.
It is confusing to get header lines, hunk headers, or context lines rendered as
being included in a custom patch, when including these makes no difference to
the patch.
This is only a visual change; internally, we still record these non-patch lines
as being included in the patch. It doesn't matter though; you can press space on
a header line and nothing happens.
It would probably be cleaner to only record + and - lines in the includedLines
array, but that would be a bit more work, and doesn't seem worth it.
I took the set of enabled checks from revive's recommended configuration [1],
and removed some that I didn't like. There might be other useful checks in
revive that we might want to enable, but this is a nice improvement already.
The bulk of the changes here are removing unnecessary else statements after
returns, but there are a few others too.
[1] https://github.com/mgechev/revive?tab=readme-ov-file#recommended-configuration
The function would return "head/branchname" when there was either a tag or a
remote with the same name.
While fixing this we slightly change the semantic of the function (and of
determineCheckedOutBranchName, which calls it): for a detached head it now
returns an empty string rather than the commit hash. I actually think this is
better.
This would crash with an index-out-of-range error.
I double-checked that all other callers of PrepareInteractiveRebaseCommand
already call getBaseHashOrRoot, so this was the only one that was broken.
I decided not to add a test for this as the scenario is not a very common one.
Pretty basic fix, didn't seem to have any complications.
I only added the refs/ prefix to the FullRefName() method
to align with other similar methods, and to make this change
not impact any user facing modals.
Fixes: https://github.com/jesseduffield/lazygit/issues/4634
Also adds a test demonstrating that the stash show behavior is now fixed
In the unlikely scenario that you have a remote branch on `origin` called
`foo`, and a local tag called `origin/foo`, git changes the behavior of
the previous command such that it produces
```
$ git for-each-ref --sort=refname --format=%(refname:short) refs/remotes
origin/branch1
remotes/origin/foo
```
with `remotes/` prepended. Presumably this is to disambiguate it from
the local tag `origin/foo`. Unfortunately, this breaks the existing
behavior of this function, so the remote branch is never shown.
By changing the command, we now get
```
$ git for-each-ref --sort=refname --format=%(refname) refs/remotes
refs/remotes/origin/branch1
refs/remotes/origin/foo
```
This allows easy parsing based on the `/`, and none of the code outside
this function has to change.
----
We previously were not showing remote HEADs for modern git versions
based on how they were formatted from "%(refname:short)".
We have decided that this is a feature, not a bug, so we are building
that into the code here.
Previously we would enter a newline at the password prompt, which would cause
the fetch to fail. The problem with this was that if you have many remotes, the
fetch would sometimes hang for some reason; I don't totally understand how that
happened, but I guess the many ssh processes requesting passwords would somehow
interfere with each other. Avoid this by simply killing the git fetch process
the moment it requests the first password.
This is a regression introduced with a199ed1396c; it is important to use a PTY
even with credentialStrategy=FAIL, otherwise the fetch command will spew the
credentials request into the UI and then hang.
This fixes the problem that background fetching makes lazygit hang when the
fetch request needs to prompt for a passphrase. For Mac users who use the
keychain to store their ssh passphrases, this can happen when lazygit is running
while the machine goes to sleep, because macOS looks the keychain in that case.
This now allows for leaving the status panel and returning back to the
same log command. Previously any return to the status panel would result
in the next command in the list being shown. Now, you need to press `a`,
with a log command being rendered, to rotate to the next
allBranchesLogCmd.
BeginInteractiveRebaseForCommit is used for all the patch commands, and for
rewording. It works by setting the commit we want to stop at to 'edit'; this
doesn't work for merge commits. This wasn't a problem for the patch commands so
far, because you typically don't use custom patches with merge commits (although
we don't prevent this; maybe we should?).
However, it was a problem when you tried to reword a merge commit; this
previously failed with an error, as the test added in the previous commit
demonstrated.
Also, we want to add a new patch command that has to stop *before* the selected
commit (pull patch to new commit before the original one), and this wouldn't
work for the first commit in a feature branch, because it would have to set the
last commit before that to 'edit', which isn't possible if that's a merge (which
is likely).
To fix all this, use a 'break' before the selected commit if the commit is a
merge. It is important that we only do it in that case and not always, otherwise
we would break the new regression tests that were added a few commits ago.
This decouples StreamOutput from whether a PTY is used. In most cases we just
want to see the output in the log window, but don't have to use a PTY, e.g. for
the bisect commands.
This has the implication that custom commands that are using "stream: true" no
longer use a PTY. In most cases that's probably a good thing, but we're going to
add a separate pty config for those who really wanted this.
Change the base type of some of our enums from int to uint8, and reorder fields
for better packing. This reduces the size of models.Commit from 152 to 132 bytes
on my machine.
This doesn't improve overall memory usage significantly, but why not save a
little bit of memory if it's easy.
We need to pass %P instead of %p in the format string of the git log command, so
that the parent hashes have the full length and can be shared with the real
hashes.
This in itself is not an improvement, because hashes are unique (they are shared
between real commits and rebase todos, but there are so few of those that it
doesn't matter). However, it becomes an improvement once we also store parent
hashes in the same pool; but the real motivation for this change is to also
reuse the hash pointers in Pipe objects later in the branch. This will be a big
win because in a merge-heavy git repo there are many more Pipe instances than
commits.
Fish shell does not support "&&" and "||" operators like
POSIX-compatible shells. Instead, it uses a different syntax structure
based on begin/end and if/else.
This caused existing lazygit nvim-remote integration templates to break
when fish was the user's default shell.
This commit adds explicit fish shell detection using the FISH_VERSION
environment variable, and provides fish-compatible templates that
correctly handle launching Neovim or sending remote commands via $NVIM.
Fixes behavior where edits would not open in a new Neovim tab or line
navigation would fail when $NVIM was set.
Ensures smoother editing experience for users running fish shell
(supported since Nov 2012 with FISH_VERSION).
We do this because
- it's closer to what you would do on the command line
- it simplifies the code a bit
- it will allow us to support cherry-picking merge commits.
What happens here is that when stopping on an "edit" todo entry, we rely on the
assumption that if the .git/rebase-merge/amend file exists, the command was
successful, and if it doesn't, there was a conflict. The problem is that when
you stop on an edit command, and then run a multi-commit cherry-pick or rebase,
this will delete the amend file. You may or may not consider this a bug in git;
to work around it, we also check the existence of the rebase-merge/message file,
which will be deleted as well by the cherry-pick or revert.
MergeRebasingCommits already merges the rebasing commits into the commits slice
that is passed in, so it doesn't make sense to append the result to commits
again. It isn't a problem, but only because commits is always empty.
It is useful to see if the conflicted commit was a "pick" or an "edit". What's
more, we're about to add support for showing cherry-picks and reverts, and
seeing that a conflicted commit was a revert is important because its diff is
backwards compared to the diff of the conflicting files in the Files panel.
This is equivalent in the current state of the code, but it will no longer be
after the next commit, because we will introduce a new status value
StatusConflicted. And in a later PR we might add yet another value
StatusCherryPicking to distinguish rebase todos from cherry-pick todos; using
commit.IsTODO is a safer way to check whether a commit is any of these.
It looks like enums.go was supposed to be file that collects a bunch of enums,
but actually there's only one in there, and since it has methods, it deserves to
be in a file of its own, named after the type.
- Remove REBASE_MODE_NORMAL. It is not the "normal" mode anyway, rather a legacy
mode; we have removed support for it in eb0f7e3d02, so there's no point in
representing it in the enum.
- Remove distinction between REBASE_MODE_REBASING and REBASE_MODE_INTERACTIVE;
these are the same now.
- Rename StatusCommands.IsInInteractiveRebase to IsInRebase.
- Remove StatusCommands.RebaseMode; use StatusCommands.IsInRebase instead.
All test cases set it to enums.REBASE_MODE_NONE, so we can simplify things a
little bit by hard-coding that. This makes the changes in the following commits
a little easier.
This makes it possible to use date and time in initial values like this:
```yaml
initialValue: 'ruudk/{{ runCommand "date +\"%Y/%-m\"" }}/'
```
I want to use this to configure my BranchPrefix like this:
```yaml
git:
branchPrefix: 'ruudk/{{ runCommand "date +\"%Y/%-m\"" }}/'
```
For the less common conflict types DD, AU, UA, DU, and UD, we would previously
only show "* Unmerged path" in the main view, which isn't helpful. Also, for
some of these we would split the main view and show this text both in the
unstaged changes and staged changes views, which is a bit embarrassing.
Improve this by offering more explanation about what's going on, and what the
most likely way to resolve the situation is for each case.
The actual usage is case insensitive, so this doesn't actually matter.
But if fills my heart with joy.
The test is case sensitive, but the actual response to `git config
commit.gpgSign` is equivalent to `git config commit.gppsign`
Add verify flag
Add and update integration tests
Rename verify to forceSkipHooks
Adapt CommitSkipHooks integration test to actually use a hook
Remove forceSkipHooks param from OnConfirm et al
Simplify tests
This is useful for copying the entire content of the selected file as it was at
the selected commit.
We only add it to the commit files panel; it is not needed in the files panel,
because there you can simply press "e" to edit the file.
This allows changing git config values while lazygit is running (e.g. in a
different terminal tab, or even in lazygit's ":" shell prompt), and have them
take effect immediately, while still getting some benefit from caching them
while lazygit is in the foreground.
Some operating systems 'open' implementations do not like
when some special characters are unencoded, so they will
double-enconde the branch name, which we already encode.
This particularly matters since branch names with / are common
For some reason the //nolint:golint,unused comment no longer seems to work after
I updated gopls (I think).
But I also don't understand why we mark unused stuff with linter comments
instead of just getting rid of it.
Some people have post-checkout hooks that take a lot of time, which makes
discarding changes slow. You can argue that a post-checkout hook should only run
when you switch branches, so it doesnt't have to run when checking out single
files or directories. You can also argue that lazygit might have implemented
discarding changes by taking the current patch and applying it in reverse, which
wouldn't have run a post-checkout hook either.
So disable them for all cases where we use git checkout with a path; this
includes checking out a file from the commit files view.
This handles the situation where the user's own config says to not show
untracked files, as is often the case with bare repos managing a user's
dotfiles.
When pushing a branch that didn't have an upstream yet, we use the command line
git push --set-upstream origin HEAD:branch-name
The HEAD: part of this is too unspecific; when checking out a different branch
while the push is still running, then git will set the upstream branch on the
newly checked out branch, not the branch that was being pushed. This might be
considered a bug in git; you might expect that it resolves HEAD at the beginning
of the operation, and uses the result at the end.
But we can easily work around this by explicitly supplying the real branch name
instead of HEAD.
We use an interactive shell so that users can use their custom shell aliases in
lazygit's shell prompt, which is convenient; however, this only really works for
shells like bash or zsh. We know it doesn't work for fish or nushell (because
these use different names for the $? variable); so use an interactive shell only
if the user's shell is either bash or zsh.
There are two ways to jump to the editor on a specific line: pressing `e` in the
staging or patch building panels, or clicking on a hyperlink in a delta diff. In
both cases, this works perfectly in the unstaged changes view, but in other
views (either staged changes, or an older commit) it can often jump to the wrong
line; this happens when there are further changes to the file being viewed in
later commits or in unstaged changes.
This commit fixes this so that you end up on the right line in these cases.
In 5a3049485c we changed the execution of shell commands to use an interactive
shell (-i), because this allows users to use aliases or shell functions, which
is a nice convenience.
Since then, however, many users have reported problems with lazygit not coming
back to the foreground after executing a shell command. Some users report that
appending "; exit" to the end of the command line solves this. I don't really
understand what the cause of this problem was, or why appending "; exit" solves
it, but if it helps, let's do it.
For non-merge commits we change "pick" to "drop" when we delete them. We do this
so that we can use the same code for dropping a commit no matter whether we are
in an interactive rebase or not. (If we aren't, we could just as well delete the
pick line from the todo list instead of setting it to "drop", but if we are, it
is better to keep the line around so that the user can change it back to "pick"
if they change their mind.)
However, merge commits can't be changed to "drop", so we have to delete them
from the todo file. We add a new daemon instruction that does this.
We still don't allow deleting a merge commit from within an interactive rebase.
The reason is that we don't show the "label" and "reset" todos in lazygit, so
deleting a merge commit would leave the commits from the branch that is being
merged in the list as "pick" commits, with no indication that they are going to
be dropped because they are on a different branch, and the merge commit that
would have brought them in is gone. This could be very confusing.
One of the comments we are deleting here said:
// Comparing just the hash is not enough; we need to compare both the
// action and the hash, as the hash could appear multiple times (e.g. in a
// pick and later in a merge).
I don't remember what I was thinking when I wrote this code, but it's nonsense
of course. Maybe I was thinking that the hash that appears in a "merge" todo
would be the hash of the commit that is being merged in (which would then
actually appear in an earlier pick), but it isn't, it's the hash of the merge
commit itself (so that the rebase can reuse its commit message). Which means
that hashes are unique, no need to compare the action.
When enabled, it adds "+n -m" after each file in the Files panel to show how
many lines were added and deleted, as with `git diff --numstat` on the command
line.
It is needed by both call sites of this function. This has the added benefit
that the argument doesn't unnecessarily show up in the status view when diffing
mode is on.
We will need this because under some conditions we are going to use this
function to edit a range of commits, and we can't set merge commits to "edit".
This corresponds to the code in startInteractiveRebaseWithEdit which has similar
logic.
It is a bit unfortunate that we will have these two different ways of setting
todos to edit: startInteractiveRebaseWithEdit does it after stopping in the
rebase, in the Then function of its refresh, but InteractiveRebase does it in
the daemon with a ChangeTodoActionsInstruction. It still makes sense though,
given how InteractiveRebase works.
This not only affects "edit", but also "drop", "fixup", and "squash".
Previously, when trying to use these for a range selection that includes a merge
commit, they would fail with the cryptic error message "Some todos not found in
git-rebase-todo"; now they simply exclude the merge commit. I'm not sure if one
is better or worse than the other, and we should probably simply disable the
commands when a merge commit is selected, but that's out of scope in this PR.
We allow deleting remote branches (or local and remote branches) only if *all*
selected branches have one.
We show the a warning about force-deleting as soon as at least one of the
selected branches is not fully merged.
The added test only tests a few of the most interesting cases; I didn't try to
cover the whole space of possible combinations, that would have been too much.
Currently we try to delete a branch normally, and if git returns an error and
its output contains the text "branch -D", then we prompt the user to force
delete, and try again using -D. Besides just being ugly, this has the
disadvantage that git's logic to decide whether a branch is merged is not very
good; it only considers a branch merged if it is either reachable from the
current head, or from its own upstream. In many cases I want to delete a branch
that has been merged to master, but I don't have master checked out, so the
current branch is really irrelevant, and it should rather (or in addition) check
whether the branch is reachable from one of the main branches. The problem is
that git doesn't know what those are.
But lazygit does, so make the check on our side, prompt the user if necessary,
and always use -D. This is both cleaner, and works better.
See this mailing list discussion for more:
https://lore.kernel.org/git/bf6308ce-3914-4b85-a04b-4a9716bac538@haller-berlin.de/
In practice, using path seems to work too, since Windows seems to be capable of
dealing with a path like C:/x/y instead of C:\x\y; but it's cleaner to do this
properly.
Also, use the user's shell (from the SHELL env variable) instead of bash. Both
of these together allow users to use their shell aliases or shell functions in
the interactive command prompt.
This change reduces the number of calls during application startup to
one, calling GetRepoPaths() earlier than previously and plumbing the
repoPaths struct around to achieve this end.
- Introduced a new optional user config command, allBranchesLogCmds
- When pressing 'a' in the Status view, cycle between non-empty, non-identical log commands
- There will always be at least one command to run, since allBranhesLogCmd has a default
- Update documentation & write an integration test
- Update translation string
The current behaviour when creating a new branch off of a remote branch
is to always track the branch it was created from.
For example, if a branch 'my_branch' is created off of the remote branch
'fix_crash_13', then 'my_branch' will be tracking the remote
'fix_crash_13' branch.
It is common practice to have both the local and remote branches named
the same when the local is tracking the remote one. Therefore, it is
reasonable to expect that 'my_branch' should not track the remote
'fix_crash_13' branch.
The new behaviour when creating a new branch off of a remote branch is
to track the branch it was created from only if the branch names match.
If the branch names DO NOT match then the newly created branch will not
track the remote branch it was created from.
For example, if a user creates a new branch 'fix_crash_13' off of the
remote branch 'fix_crash_13', then the local 'fix_crash_13' branch will
track the remote 'fix_crash_13' branch.
However, if the user creates a new branch called 'other_branch_name' off
of the remote branch 'fix_crash_13', then the local 'other_branch_name'
branch will NOT track the remote 'fix_crash_13' branch.
Several custom patch commands on parts of an added file would fail with the
confusing error message "error: new file XXX depends on old contents". These
were dropping the custom patch from the original commit, moving the patch to a
new commit, moving it to a later commit, or moving it to the index.
We fix this by converting the patch header from an added file to a diff against
an empty file. We do this not just for the purpose of applying the patch, but
also for rendering it and copying it to the clip board. I'm not sure it matters
much in these cases, but it does feel more correct for a filtered patch to be
presented this way.
We're going to add another argument in the next commit, and that's getting a bit
much, especially when most of the arguments are bool and you only see true and
false at the call sites without knowing what they mean.
Unfortunately it isn't possible to delete them. This would often be useful, but
our todo rewriting mechanisms rely on being able to find todos by some
identifier (hash for pick, ref for update-ref), and exec todos don't have a
unique identifier.
It is a valid case for a branch to share no history with any of the main
branches, in which case git merge-base returns an error (and an empty string).
Since we can't distinguish this from one of the main branches having been
deleted, we shouldn't invalidate the cache in that case.
This broke with #3528.
If the remote branch is not stored locally, we only see question marks in the
branch status. In this case we can't tell whether we need to force-push, so it's
best to assume that we don't, and see if the server rejects the push, and react
to that by asking to force push. This second part is also broken right now,
we'll fix this in the next commit.
To determine whether we need to ask for force pushing, we need to query the push
branch rather than the upstream branch, in case they are not the same.
In a triangular workflow the branch that you're pulling from is not the same as
the one that you are pushing to. For example, some people find it useful to set
the upstream branch to origin/master so that pulling effectively rebases onto
master, and set the push.default git config to "current" so that "feature"
pushes to origin/feature.
Another example is a fork-based workflow where "feature" has upstream set to
upstream/main, and the repo has remote.pushDefault set to "origin", so pushing
on "feature" pushes to origin/feature.
This commit adds new fields to models.Branch that store the ahead/behind
information against the push branch; for the "normal" workflow where you pull
and push from/to the upstream branch, AheadForPush/BehindForPush will be the
same as AheadForPull/BehindForPull.
In go 1.22, loop variables are redeclared with each iteration of the
loop, rather than simple updated on each iteration. This means that we
no longer need to manually redeclare variables when they're closed over
by a function.
We are going to truncate overly long lines returned from git log, and the most
likely field that is going to make the line too long is the subject; so we must
put it last, otherwise we'd end up with not enough fields to split when it's too
long.
It might not be obvious from the diff what's happening to the mock command
output in the test: it didn't have the divergence field (">") at all, which was
kind of a bug. It didn't matter for these tests though, because we are not
testing the divergence here, and our production code happens to be resilient
against it missing. But now we must add the ">" field before the subject.
Scanners can return errors (e.g. ErrTooLong), and if we don't handle it, the
cmd.Wait() call below will block forever because nobody drains the command's
output.
This happens for CommitLoader.GetCommits when there's a commit whose subject
line is longer than approx. 65500 characters; in that case, lazygit would lock
up completely. With this fix it remains usable, but the commit list is truncated
before the bad commit, which is not good enough. We'll improve that in the
remaining commits of this branch.
The rebase.updateRefs feature of git is very useful to rebase a stack of
branches and keep everything nicely stacked; however, it is usually in the way
when you make a copy of a branch and want to rebase it "away" from the original
branch in some way or other. For example, the original branch might sit on main,
and you want to rebase the copy onto devel to see if things still compile there.
Or you want to do some heavy history rewriting experiments on the copy, but keep
the original branch in case the experiments fail. Or you want to split a branch
in two because it contains two unrelated sets of changes; so you make a copy,
and drop half of the commits from the copy, then check out the original branch
and drop the other half of the commits from it.
In all these cases, git's updateRefs feature insists on moving the original
branch along with the copy in the first rebase that you make on the copy. I
think this is a bug in git, it should create update-ref todos only for branches
that point into the middle of your branch (because only then do they form a
stack), not when they point at the head (because then it's a copy). I had a long
discussion about this on the git mailing list [1], but people either don't agree
or don't care enough.
So we fix this on our side: whenever we start a rebase for whatever reason, be
it interactive, non-interactive, or behind-the-scenes, we drop any update-ref
todos that are at the very top of the todo list, which fixes all the
above-mentioned scenarios nicely.
I will admit that there's one scenario where git's behavior is the desired one,
and the fix in this PR makes it worse: when you create a new branch off of an
existing one, with the intention of creating a stack of branches, but before you
make the first commit on the new branch you realize some problem with the first
branch (e.g. a commit that needs to be reworded or dropped). It this case you do
want both branches to be affected by the change. In my experience this scenario
is much rarer than the other ones that I described above, and it's also much
easier to recover from: just check out the other branch again and hard-reset it
to the rebased one.
[1]
https://public-inbox.org/git/354f9fed-567f-42c8-9da9-148a5e223022@haller-berlin.de/
