This script is failing currently on
https://github.com/jesseduffield/lazygit/pull/3631 because that fork's
master branch is 300 commits behind our own, but the feature branch is
up to date.
The thing is, we don't actually need to involve the master branch. All
we care about is the feature branch's own commits, so this commit simply
fetches those commits and checks them.
- **PR Description**
Support per-repo user config files. For now we only support
`.git/lazygit.yml`; in the future we would also like to support
`./.lazygit.yml`, but that one will need a trust prompt as it could be
versioned, which adds quite a bit of complexity, so we leave that for
later.
We do, however, support config files in parent directories (all the way
up to the root directory). This makes it possible to add a config file
that applies to multiple repos at once. Useful if you want to set
different options for all your work repos vs. all your open-source
repos, for instance.
In addition, we support re-loading edited config files. This makes it
much easier to experiment with config settings, especially the ones that
affect the layout or color scheme, because you see the effect
immediately without having to restart lazygit.
Since onNewRepo calls resetKeybindings, which reinitializes the keybindings for
custom commands, all we have to do for this is store a pointer to a config
instead of storing the customCommands, so we get the up-to-date ones every time.
It still doesn't update the view immediately when reloading the config, only the
next time the status panel is focused. But at least it does it then; writing the
code to update the panel when reloading the config is probably not justifiable.
This allows having per-repo config files with different languages per repo. Now
granted, this is not an important use case that we need to support; however, the
goal is to eventually make all configs hot-reloadable (as opposed to loading
them only once at startup), so this is one step in that direction.
For now we only support .git/lazygit.yml; in the future we would also like to
support ./.lazygit.yml, but that one will need a trust prompt as it could be
versioned, which adds quite a bit of complexity, so we leave that for later.
We do, however, support config files in parent directories (all the way up to
the root directory). This makes it possible to add a config file that applies to
multiple repos at once. Useful if you want to set different options for all your
work repos vs. all your open-source repos, for instance.
Currently, userConfig is only read once at startup and then never changes. Later
in this branch, we will add the possibility to reload the user config; this can
happen either when switching repositories, or when the user has edited the
config file. In both cases, reloading happens on the main thread, but the user
config could be accessed concurrently from background threads, so we need to
make this safe.
At the moment, the user config is only read once at startup, so there's no point
in writing it back to disk. However, later in this branch we will add code that
reloads the user config when switching repos, which does happen quite a bit in
integration tests; this would undo the changes that a test made in its
SetupConfig function, so write those changes to disk to prevent that from
happening.
This makes it more explicit how to deal with the different types of config
files: a user-supplied config file (via the LG_CONFIG_FILE env var) is required
to exist, whereas the default config file will be created if it is missing.
We will later extend this with repo-specific config files, which will be skipped
if missing.
Split it so createAllViews instanciates the views, and sets those properties
that are independent of the user config, and configureViewProperties which sets
those things that do depend on the user config. For now we call the second right
after the first, but later we'll call configureViewProperties after reloading
the user config.
It was added in 043cb2ea44, and the commit message was "reload config whenever
returning to gui". I don't understand what this means; Run() is called exactly
once after startup, so it would just reload the config again for no reason.
We will add a real way of reloading the config whenever it has changed later in
this branch.
We are going to make a few changes to the fields in this branch, and we can make
them with more peace of mind when we can be sure they are not accessed from
outside this package.
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.
- **PR Description**
When clicking in a single-file diff view to enter staging (or custom
patch editing, when coming from the commit files panel), you needed to
press escape twice to exit, where the first press would seemingly do
nothing.
When clicking in a single-file diff view to enter staging (or custom patch
editing, when coming from the commit files panel), you needed to press escape
twice to exit, where the first press would seemingly do nothing.
The reason for this was that after clicking in the diff we end up in non-sticky
range select mode, but only with a single line selected, which is basically
indistinguishable from line select mode.
- **PR Description**
Improve the template placeholders that are available for custom
commands:
- `SelectedCommit` replaces `SelectedLocalCommit`,
`SelectedReflogCommit`, and `SelectedSubCommit`
- `SelectedPath` is set to `SelectedCommitFilePath` when the CommitFiles
context is active
It still slightly bothers me that we don't make a similar unification
for `SelectedLocalBranch` and `SelectedRemoteBranch` (and others), but
it would be a bigger change to do that, and we decided in #3663 not to.
Fixes#3663.
SelectedCommit is context-dependent and points to SelectedLocalCommit,
SelectedReflogCommit, or SelectedSubCommit depending on which panel is active.
If none of these panels is active, it returns the selected local commit, which
is probably the most useful default (e.g. when defining custom commands for the
Files panel).
The comments that I'm deleting here explain why we need the bool; however, in
our case that's a theoretical issue. It would only arise if we ever were to pass
a nil context to SetParentContext, which we never do.
