git-gui - re-enable use of hook scripts #100
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Earlier, commit aae9560 introduced search in $PATH to find executables before running them, avoiding an issue where on Windows a same named file in the current directory can be executed in preference to anything in a directory in $PATH. This search is intended to find an absolute path for a bare executable ( e.g, a function "foo") by finding the first instance of "foo" in a directory given in $PATH, and this search works correctly. The search is explicitly avoided for an executable named with an absolute path (e.g., /bin/sh), and that works as well. Unfortunately, the search is also applied to commands named with a relative path. A hook script (or executable) $HOOK is usually located relative to the project directory as .git/hooks/$HOOK. The search for this will generally fail as that relative path will (probably) not exist on any directory in $PATH. This means that git hooks in general now fail to run. Considerable mayhem could occur should a directory on $PATH be git controlled. If such a directory includes .git/hooks/$HOOK, that repository's $HOOK will be substituted for the one in the current project, with unknown consequences. This lookup failure also occurs in worktrees linked to a remote .git directory using git-new-workdir. However, a worktree using a .git file pointing to a separate git directory apparently avoids this: in that case the hook command is resolved to an absolute path before being passed down to the code introduced in aae9560. Fix this by replacing the test for an "absolute" pathname to a check for a command name having more than one pathname component. This limits the search and absolute pathname resolution to bare commands. The new test uses tcl's "file split" command. Experiments on Linux and Windows, using tclsh, show that command names with relative and absolute paths always give at least two components, while a bare command gives only one. Linux: puts [file split {foo}] ==> foo Linux: puts [file split {/foo}] ==> / foo Linux: puts [file split {.git/foo}] ==> .git foo Windows: puts [file split {foo}] ==> foo Windows: puts [file split {c:\foo}] ==> c:/ foo Windows: puts [file split {.git\foo}] ==> .git foo The above results show the new test limits search and replacement to bare commands on both Linux and Windows. Signed-off-by: Mark Levedahl <mlevedahl@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
I am planning to merge this directly to my tree (via -Xsubtree=git-gui/ option), and thought that it would make it easier to maintain your tree up to date by simply fast-forwarding to it (the commit is built directly on top of your 'master'). Thanks. |
git-gui currently runs some hooks directly using its own code written before 2010, long predating git v2.9 that added the core.hooksPath configuration to override the assumed location at $GIT_DIR/hooks. Thus, git-gui looks for and runs hooks including prepare-commit-msg, commit-msg, pre-commit, post-commit, and post-checkout from $GIT_DIR/hooks, regardless of configuration. Commands (e.g., git-merge) that git-gui invokes directly do honor core.hooksPath, meaning the overall behaviour is inconsistent. Furthermore, since v2.36 git exposes its hook execution machinery via `git-hook run`, eliminating the need for others to maintain code duplicating that functionality. Using git-hook will both fix git-gui's current issues on hook configuration and (presumably) reduce the maintenance burden going forward. So, teach git-gui to use git-hook. Signed-off-by: Mark Levedahl <mlevedahl@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
|
Hi @gitster I fetched your tree just now and I see that you have not yet merged it. My |
|
My copy already has these in 'next', but they are identical so it is not a huge deal. |
ghost
approved these changes
Sep 28, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Earlier, commit aae9560 introduced search in $PATH to find executables before running them, avoiding an issue where on Windows a same named file in the current directory can be executed in preference to anything in a directory in $PATH. This search is intended to find an absolute path for a bare executable ( e.g, a function "foo") by finding the first instance of "foo" in a directory given in $PATH, and this search works correctly. The search is explicitly avoided for an executable named with an absolute path (e.g., /bin/sh), and that works as well.
Unfortunately, the search is also applied to commands named with a relative path. A hook script (or executable) $HOOK is usually located relative to the project directory as .git/hooks/$HOOK. The search for this will generally fail as that relative path will (probably) not exist on any directory in $PATH. This means that git hooks in general now fail to run. Considerable mayhem could occur should a directory on $PATH be git controlled. If such a directory includes .git/hooks/$HOOK, that repository's $HOOK will be substituted for the one in the current project, with unknown consequences.
This lookup failure also occurs in worktrees linked to a remote .git directory using git-new-workdir. However, a worktree using a .git file pointing to a separate git directory apparently avoids this: in that case the hook command is resolved to an absolute path before being passed down to the code introduced in aae9560.
Fix this by replacing the test for an "absolute" pathname to a check for a command name having more than one pathname component. This limits the search and absolute pathname resolution to bare commands. The new test uses tcl's "file split" command. Experiments on Linux and Windows, using tclsh, show that command names with relative and absolute paths always give at least two components, while a bare command gives only one.
The above results show the new test limits search and replacement to bare commands on both Linux and Windows.