git-gui: respect core.hooksPath, falling back to .git/hooks

From: Johannes Schindelin <[email protected]>

Since v2.9.0, Git knows about the config variable core.hookspath that allows overriding the path to the directory containing the Git hooks.

Since v2.10.0, the --git-path option respects that config variable, too, so we may just as well use that command.

For Git versions older than v2.5.0 (which was the first version to support the --git-path option for the rev-parse command), we simply fall back to the previous code.

This fixes https://github.com/git-for-windows/git/issues/1755

Initial-patch-by Philipp Gortan <[email protected]>

Signed-off-by Johannes Schindelin <[email protected]>

---
 git-gui.sh | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

git-gui.sh +0 -0

index fd476b6999..b2c6e7a1db 100755

Hi,

Quoted 28 lines · Johannes Schindelin via GitGitGadget

From: Johannes Schindelin <[email protected]>

Since v2.9.0, Git knows about the config variable core.hookspath that allows overriding the path to the directory containing the Git hooks.

Since v2.10.0, the --git-path option respects that config variable, too, so we may just as well use that command.

For Git versions older than v2.5.0 (which was the first version to support the --git-path option for the rev-parse command), we simply fall back to the previous code.

This fixes https://github.com/git-for-windows/git/issues/1755

Initial-patch-by Philipp Gortan <[email protected]>

Signed-off-by Johannes Schindelin <[email protected]>

---
 git-gui.sh | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

git-gui.sh +0 -0

index fd476b6999..b2c6e7a1db 100755
@@ -623,7 +623,11 @@ proc git_write {args} {
 }
 

git-gui.sh proc git_write {args} { · line 630

Quoted 7 lines

git-gui.sh +5 -1

 proc githook_read {hook_name args} {
-     set pchook [gitdir hooks $hook_name]
+     if {[package vcompare $::_git_version 2.5.0] >= 0} {
+             set pchook [git rev-parse --git-path "hooks/$hook_name"]
+     } else {
+             set pchook [gitdir hooks $hook_name]
+     }

gitdir is used in a lot of places, and I think all those would also benefit from using --git-path. So I think it is a better idea to move this to the procedure gitdir. It would have to be refactored to take any number of arguments, instead of the two it takes here.

Other than that, looks good. Thanks.

--
Regards,
Pratyush Yadav

Quoted 32 lines · Pratyush Yadav

Hi,

Quoted 28 lines · Johannes Schindelin via GitGitGadget

From: Johannes Schindelin <[email protected]>

Since v2.9.0, Git knows about the config variable core.hookspath that allows overriding the path to the directory containing the Git hooks.

Since v2.10.0, the --git-path option respects that config variable, too, so we may just as well use that command.

For Git versions older than v2.5.0 (which was the first version to support the --git-path option for the rev-parse command), we simply fall back to the previous code.

This fixes https://github.com/git-for-windows/git/issues/1755

Initial-patch-by Philipp Gortan <[email protected]>

Signed-off-by Johannes Schindelin <[email protected]>

---
 git-gui.sh | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

git-gui.sh +0 -0

index fd476b6999..b2c6e7a1db 100755
@@ -623,7 +623,11 @@ proc git_write {args} {
 }
 

Quoted 12 lines

git-gui.sh proc git_write {args} { · line 630

Quoted 7 lines

git-gui.sh +5 -1

 proc githook_read {hook_name args} {
-     set pchook [gitdir hooks $hook_name]
+     if {[package vcompare $::_git_version 2.5.0] >= 0} {
+             set pchook [git rev-parse --git-path "hooks/$hook_name"]
+     } else {
+             set pchook [gitdir hooks $hook_name]
+     }

gitdir is used in a lot of places, and I think all those would also benefit from using --git-path. So I think it is a better idea to move this to the procedure gitdir. It would have to be refactored to take any number of arguments, instead of the two it takes here.

gitdir already takes an arbitrary number of arguments and joins them to a path. The more imminent challenge is, that gitdir caches the GIT_DIR, thus it tries to avoid calling "git rev-parse". Which works for most, but not for hooks.

We could either maintain a blacklist, for what we cache the result too, or always call "git rev-parse --git-dir".

This blacklist would need to be in sync with the one in Git's path.c::adjust_git_path() than.

Quoted 44 lines · Bert Wesarg

Quoted 30 lines · Pratyush Yadav

Quoted 28 lines · Johannes Schindelin via GitGitGadget

From: Johannes Schindelin <[email protected]>

Since v2.9.0, Git knows about the config variable core.hookspath that allows overriding the path to the directory containing the Git hooks.

Since v2.10.0, the --git-path option respects that config variable, too, so we may just as well use that command.

For Git versions older than v2.5.0 (which was the first version to support the --git-path option for the rev-parse command), we simply fall back to the previous code.

This fixes https://github.com/git-for-windows/git/issues/1755

Initial-patch-by Philipp Gortan <[email protected]>

Signed-off-by Johannes Schindelin <[email protected]>

---
 git-gui.sh | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

git-gui.sh +0 -0

index fd476b6999..b2c6e7a1db 100755
@@ -623,7 +623,11 @@ proc git_write {args} {
 }
 

Quoted 12 lines

git-gui.sh proc git_write {args} { · line 630

Quoted 7 lines

git-gui.sh +5 -1

 proc githook_read {hook_name args} {
-     set pchook [gitdir hooks $hook_name]
+     if {[package vcompare $::_git_version 2.5.0] >= 0} {
+             set pchook [git rev-parse --git-path "hooks/$hook_name"]
+     } else {
+             set pchook [gitdir hooks $hook_name]
+     }

gitdir is used in a lot of places, and I think all those would also benefit from using --git-path. So I think it is a better idea to move this to the procedure gitdir. It would have to be refactored to take any number of arguments, instead of the two it takes here.

Quoted 4 lines

gitdir already takes an arbitrary number of arguments and joins them to a path. The more imminent challenge is, that gitdir caches the GIT_DIR, thus it tries to avoid calling "git rev-parse". Which works for most, but not for hooks.

What I was thinking of was something like this:

- If no args are passed, then just directly return $_gitdir. This is already being done. I assume the GIT_DIR relocation is already handled by git rev-parse --git-dir, so this would point to the correct location.
- If args are passed, then we want a subdirectory of GIT_DIR In this case, it is possible that this subdirectory has also been relocated (hooks/ being one of those subdirectories). So in this case, use git rev-parse --git-path instead.

So the gitdir procedure would look something like:

  proc gitdir {args} {
  	global $_gitdir
  	if {$args eq {}} {
  		# Return the cached GIT_DIR
  		return $_gitdir
  	}

  	# Use `git rev-parse --git-path` to get the path instead of
  	# using the cached value.
  }

Am I missing something? Or does this fix the issue you describe?

Quoted 5 lines

We could either maintain a blacklist, for what we cache the result too, or always call "git rev-parse --git-dir".

This blacklist would need to be in sync with the one in Git's path.c::adjust_git_path() than.

Is caching GIT_DIR that important in terms of performance? Otherwise, I'd say calling git rev-parse --git-path for _every_ subdirectory of GIT_DIR is a much simpler solution.

--
Regards,
Pratyush Yadav

Quoted 7 lines · Pratyush Yadav

Quoted 6 lines · Bert Wesarg

Quoted 5 lines · Pratyush Yadav

gitdir is used in a lot of places, and I think all those would also
benefit from using --git-path. So I think it is a better idea to move this to the procedure gitdir. It would have to be refactored to take any number of arguments, instead of the two it takes here.

The gitdir function is called 13 times during startup alone, and who knows how many more times later.

So I am quite convinced that the original intention was to save on spawning processes left and right.

But since you are the Git GUI maintainer, and this was your suggestion, I made it so.

Quoted 5 lines

Quoted 5 lines

We could either maintain a blacklist, for what we cache the result too, or always call "git rev-parse --git-dir".

This blacklist would need to be in sync with the one in Git's path.c::adjust_git_path() than.

Quoted 9 lines · Johannes Schindelin

Quoted 7 lines · Pratyush Yadav

Quoted 6 lines · Bert Wesarg

Quoted 5 lines · Pratyush Yadav

gitdir is used in a lot of places, and I think all those would also
benefit from using --git-path. So I think it is a better idea to move this to the procedure gitdir. It would have to be refactored to take any number of arguments, instead of the two it takes here.

Quoted 8 lines

The gitdir function is called 13 times during startup alone, and who knows how many more times later.

So I am quite convinced that the original intention was to save on spawning processes left and right.

But since you are the Git GUI maintainer, and this was your suggestion, I made it so.

Yes, I am the maintainer, but I am not an all-knowing, all-seeing entity. Your, and every other contributors, suggestions are very valuable. And my suggestions aren't gospel. I would hate to see someone send in a patch they weren't sure was the best thing to do just because I suggested it. Please feel free to object my suggestions.

In this case, I didn't expect gitdir to be called this many times.

While I don't notice much of a performance difference on my system (Linux), a quick measurement tells me that the time spent in gitdir is about 16 ms. In contrast, the same measurement without the v2 patch gives out 0 ms (IOW, very fast). 16 ms sounds a bit much for something so simple. It might not be the same for everyone else. AFAIK, spawning a process is much slower on Windows.

So now I'm not so sure my suggestion was a good one. My original aim was to be sure everything was correct, and no incorrect directories were being used. But the current solution comes at a performance hit.

Quoted 3 lines

Quoted 3 lines

Quoted 3 lines

We could either maintain a blacklist, for what we cache the result too, or always call "git rev-parse --git-dir".

Quoted 2 lines

Quoted 2 lines

Quoted 2 lines

This blacklist would need to be in sync with the one in Git's path.c::adjust_git_path() than.

Bert's suggestion seems like a decent compromise. We run git rev-parse --git-path for the paths in the blacklist, and for the rest we use the cached value. This does run the risk of getting out of sync with git.git's list, but it might be better than spawing a process every time, and is very likely better than just doing it for hooks.

Hi,

Quoted 41 lines · Pratyush Yadav

Quoted 9 lines · Johannes Schindelin

Quoted 7 lines · Pratyush Yadav

Quoted 6 lines · Bert Wesarg

Quoted 5 lines · Pratyush Yadav

gitdir is used in a lot of places, and I think all those would also
benefit from using --git-path. So I think it is a better idea to move this to the procedure gitdir. It would have to be refactored to take any number of arguments, instead of the two it takes here.

Quoted 8 lines

The gitdir function is called 13 times during startup alone, and who knows how many more times later.

So I am quite convinced that the original intention was to save on spawning processes left and right.

But since you are the Git GUI maintainer, and this was your suggestion, I made it so.

Yes, I am the maintainer, but I am not an all-knowing, all-seeing entity. Your, and every other contributors, suggestions are very valuable. And my suggestions aren't gospel. I would hate to see someone send in a patch they weren't sure was the best thing to do just because I suggested it. Please feel free to object my suggestions.

In this case, I didn't expect gitdir to be called this many times.

While I don't notice much of a performance difference on my system (Linux), a quick measurement tells me that the time spent in gitdir is about 16 ms. In contrast, the same measurement without the v2 patch gives out 0 ms (IOW, very fast). 16 ms sounds a bit much for something so simple. It might not be the same for everyone else. AFAIK, spawning a process is much slower on Windows.

So now I'm not so sure my suggestion was a good one. My original aim was to be sure everything was correct, and no incorrect directories were being used. But the current solution comes at a performance hit.

Quoted 3 lines

Quoted 3 lines

Quoted 3 lines

We could either maintain a blacklist, for what we cache the result too, or always call "git rev-parse --git-dir".

Quoted 8 lines

Quoted 2 lines

Quoted 2 lines

Quoted 2 lines

This blacklist would need to be in sync with the one in Git's path.c::adjust_git_path() than.

Bert's suggestion seems like a decent compromise. We run git rev-parse --git-path for the paths in the blacklist, and for the rest we use the cached value. This does run the risk of getting out of sync with git.git's list, but it might be better than spawing a process every time, and is very likely better than just doing it for hooks.

But what about this part of that function?

-- snip --
else if (repo->different_commondir)

	update_common_dir(buf, git_dir_len, repo->commondir);

-- snap --

It might well turn out that this blacklist is neither easy to implement nor will it help much.

So let's look at all the call sites:

-- snip --

$ git grep -w gitdir | sed -ne 's|\].*||' -e 's|.*\[gitdir ||p' | sort | uniq
$file
$name
CHERRY_PICK_HEAD
FETCH_HEAD
GITGUI_BCK
GITGUI_EDITMSG
GITGUI_MSG
HEAD
hooks $hook_name
index.lock
info exclude
logs $name
MERGE_HEAD
MERGE_MSG
MERGE_RR
objects 4\[0-[expr {$ndirs-1}
objects info
objects info alternates
objects pack
packed-refs
PREPARE_COMMIT_MSG
rebase-merge head-name
remotes
remotes $r
rr-cache
rr-cache MERGE_RR
SQUASH_MSG
-- snap --

The $file call looks for messages (probably commit, merge, tag messages and the likes), the $name one looks for refs.

Some of those arguments strike me as very good candidates to require the common directory while others require the real gitdir (remember, commondir != gitdir in worktrees other than the main worktree).

What _could_ be done (but we're certainly threatening to enter the realm of the ridiculous here) is to call git rev-parse --git-dir --git-path CHERRY_PICK_HEAD --git-path FETCH_HEAD [...], which will output one path per line, and then store the result in an associative array
(https://tcl.tk/man/tcl8.5/tutorial/Tcl22.html), and use that to look up paths based on their first component, caching as we go.

Something like this:

-- snipsnap --

git-gui.sh +41 -2

index fd476b6..9295c75 100755
@@ -158,6 +158,7 @@ if {[tk windowingsystem] eq "aqua"} {
 
 set _appname {Git Gui}
 set _gitdir {}
+array set _gitdir_cached {}
 set _gitworktree {}
 set _isbare {}
 set _gitexec {}

@@ -197,12 +198,50 @@ proc appname {} {
 	return $_appname
 }
 
+proc init_gitdir_cached {} {
+	global _gitdir _gitdir_cached
+
+	set gitdir_keys [list \
+		CHERRY_PICK_HEAD FETCH_HEAD GITGUI_BCK GITGUI_EDITMSG \
+		GITGUI_MSG HEAD hooks index.lock info logs MERGE_HEAD \
+		MERGE_MSG MERGE_RR objects packed-refs PREPARE_COMMIT_MSG \
+		rebase-merge head-name remotes rr-cache SQUASH_MSG \
+		]
+
+	set gitdir_cmd [list git rev-parse --git-dir]
+	foreach key $gitdir_keys {
+		lappend gitdir_cmd --git-path $key
+	}
+
+	set i -1
+	foreach path [split [eval $gitdir_cmd] "\n"] {
+		if {$i eq -1} {
+			set _gitdir $path
+		} else {
+			set _gitdir_cached([lindex $gitdir_keys $i]) $path
+		}
+		incr i
+	}
+}
+
 proc gitdir {args} {
-	global _gitdir
+	global _gitdir _gitdir_cached
+
 	if {$args eq {}} {
 		return $_gitdir
 	}
-	return [eval [list file join $_gitdir] $args]
+
+	set arg0 [lindex $args 0]
+	set args [lrange $args 1 end]
+	if {![info exists _gitdir_cached($arg0)]} {
+		if {[package vcompare $::_git_version 2.5.0] >= 0} {
+			set _gitdir_cached($arg0) [git rev-parse --git-path $arg0]
+		} else {
+			set _gitdir_cached($arg0) [file join $_gitdir $arg0]
+		}
+	}
+
+	return [eval [concat [list file join $_gitdir_cached($arg0)] $args]]
 }
 
 proc gitexec {args} {

Quoted 53 lines · Johannes Schindelin

Quoted 41 lines · Pratyush Yadav

Quoted 9 lines · Johannes Schindelin

Quoted 7 lines · Pratyush Yadav

Quoted 6 lines · Bert Wesarg

Quoted 5 lines · Pratyush Yadav

gitdir is used in a lot of places, and I think all those would also
benefit from using --git-path. So I think it is a better idea to move this to the procedure gitdir. It would have to be refactored to take any number of arguments, instead of the two it takes here.

Quoted 8 lines

The gitdir function is called 13 times during startup alone, and who knows how many more times later.

So I am quite convinced that the original intention was to save on spawning processes left and right.

But since you are the Git GUI maintainer, and this was your suggestion, I made it so.

Yes, I am the maintainer, but I am not an all-knowing, all-seeing entity. Your, and every other contributors, suggestions are very valuable. And my suggestions aren't gospel. I would hate to see someone send in a patch they weren't sure was the best thing to do just because I suggested it. Please feel free to object my suggestions.

In this case, I didn't expect gitdir to be called this many times.

While I don't notice much of a performance difference on my system (Linux), a quick measurement tells me that the time spent in gitdir is about 16 ms. In contrast, the same measurement without the v2 patch gives out 0 ms (IOW, very fast). 16 ms sounds a bit much for something so simple. It might not be the same for everyone else. AFAIK, spawning a process is much slower on Windows.

So now I'm not so sure my suggestion was a good one. My original aim was to be sure everything was correct, and no incorrect directories were being used. But the current solution comes at a performance hit.

Quoted 3 lines

Quoted 3 lines

Quoted 3 lines

We could either maintain a blacklist, for what we cache the result too, or always call "git rev-parse --git-dir".

Quoted 8 lines

Quoted 2 lines

Quoted 2 lines

Quoted 2 lines

This blacklist would need to be in sync with the one in Git's path.c::adjust_git_path() than.

Bert's suggestion seems like a decent compromise. We run git rev-parse --git-path for the paths in the blacklist, and for the rest we use the cached value. This does run the risk of getting out of sync with git.git's list, but it might be better than spawing a process every time, and is very likely better than just doing it for hooks.

Quoted 6 lines

But what about this part of that function?

-- snip --
else if (repo->different_commondir)

	update_common_dir(buf, git_dir_len, repo->commondir);

-- snap --

I'm afraid I'm a bit out of my depth on this. I have no idea what a "common directory" is, and how is it different from the "git directory". I can't find anything useful on Google about it. My guess is that it is something related to separate worktrees.

Quoted 2 lines

It might well turn out that this blacklist is neither easy to implement nor will it help much.

Am I correct in assuming that for other cases like "info", "grafts", "index", "objects", and "hooks" the blacklist would be simple to implement, and it is the "common directory" case that is problematic?

Quoted 33 lines

So let's look at all the call sites:

-- snip --

$ git grep -w gitdir | sed -ne 's|\].*||' -e 's|.*\[gitdir ||p' | sort | uniq
$file
$name
CHERRY_PICK_HEAD
FETCH_HEAD
GITGUI_BCK
GITGUI_EDITMSG
GITGUI_MSG
HEAD
hooks $hook_name
index.lock
info exclude
logs $name
MERGE_HEAD
MERGE_MSG
MERGE_RR
objects 4\[0-[expr {$ndirs-1}
objects info
objects info alternates
objects pack
packed-refs
PREPARE_COMMIT_MSG
rebase-merge head-name
remotes
remotes $r
rr-cache
rr-cache MERGE_RR
SQUASH_MSG
-- snap --

Quoted 2 lines

The $file call looks for messages (probably commit, merge, tag messages and the likes), the $name one looks for refs.

So they should always be inside the '.git' or GIT_DIR, correct?

Quoted 10 lines

Some of those arguments strike me as very good candidates to require the common directory while others require the real gitdir (remember, commondir != gitdir in worktrees other than the main worktree).

What _could_ be done (but we're certainly threatening to enter the realm of the ridiculous here) is to call git rev-parse --git-dir --git-path CHERRY_PICK_HEAD --git-path FETCH_HEAD [...], which will output one path per line, and then store the result in an associative array
(https://tcl.tk/man/tcl8.5/tutorial/Tcl22.html), and use that to look up paths based on their first component, caching as we go.

Ah yes! That is certainly threatening to enter the realm of ridiculous. I'm not sure what benefit this will have. Right now, I don't think git-gui handles these cases. Have people complained? Is this a common problem?

I want to evaluate how much benefit we get doing something like this has over just using your original patch that works with hooks only.

Quoted 67 lines

Something like this:

-- snipsnap --

git-gui.sh +41 -2

index fd476b6..9295c75 100755
@@ -158,6 +158,7 @@ if {[tk windowingsystem] eq "aqua"} {
 
 set _appname {Git Gui}
 set _gitdir {}
+array set _gitdir_cached {}
 set _gitworktree {}
 set _isbare {}
 set _gitexec {}

@@ -197,12 +198,50 @@ proc appname {} {
 	return $_appname
 }
 
+proc init_gitdir_cached {} {
+	global _gitdir _gitdir_cached
+
+	set gitdir_keys [list \
+		CHERRY_PICK_HEAD FETCH_HEAD GITGUI_BCK GITGUI_EDITMSG \
+		GITGUI_MSG HEAD hooks index.lock info logs MERGE_HEAD \
+		MERGE_MSG MERGE_RR objects packed-refs PREPARE_COMMIT_MSG \
+		rebase-merge head-name remotes rr-cache SQUASH_MSG \
+		]
+
+	set gitdir_cmd [list git rev-parse --git-dir]
+	foreach key $gitdir_keys {
+		lappend gitdir_cmd --git-path $key
+	}
+
+	set i -1
+	foreach path [split [eval $gitdir_cmd] "\n"] {
+		if {$i eq -1} {
+			set _gitdir $path
+		} else {
+			set _gitdir_cached([lindex $gitdir_keys $i]) $path
+		}
+		incr i
+	}
+}
+
 proc gitdir {args} {
-	global _gitdir
+	global _gitdir _gitdir_cached
+
 	if {$args eq {}} {
 		return $_gitdir
 	}
-	return [eval [list file join $_gitdir] $args]
+
+	set arg0 [lindex $args 0]
+	set args [lrange $args 1 end]
+	if {![info exists _gitdir_cached($arg0)]} {
+		if {[package vcompare $::_git_version 2.5.0] >= 0} {
+			set _gitdir_cached($arg0) [git rev-parse --git-path $arg0]
+		} else {
+			set _gitdir_cached($arg0) [file join $_gitdir $arg0]
+		}
+	}
+
+	return [eval [concat [list file join $_gitdir_cached($arg0)] $args]]
 }
 

git-gui.sh if {[catch { · line 1287

Quoted 10 lines

git-gui.sh +1 -1

 proc gitexec {args} {

@@ -1242,7 +1281,7 @@ if {[catch {
 	&& [catch {
 		# beware that from the .git dir this sets _gitdir to .
 		# and _prefix to the empty string
-		set _gitdir [git rev-parse --git-dir]
+		init_gitdir_cached
 		set _prefix [git rev-parse --show-prefix]
 	} err]} {
 	load_config 1

A nice way of tackling this problem overall considering the challenges, but I'm worried about whether all this is _actually_ needed for real use cases, and what breaks if we don't.

Honestly, I'm not too sure how to tackle this problem. That is also the reason I took so long in writing this response. What would your suggestion be?

Also, if some other people interested in git-gui could chime in, it would be great.

From: Johannes Schindelin <[email protected]>

Since v2.9.0, Git knows about the config variable core.hookspath that allows overriding the path to the directory containing the Git hooks.

Since v2.10.0, the --git-path option respects that config variable, too, so we may just as well use that command.

For Git versions older than v2.5.0 (which was the first version to support the --git-path option for the rev-parse command), we simply fall back to the previous code.

An original patch handled only the hooksPath setting, however, during the code submission it was deemed better to fix all call to the gitdir function.

To avoid spawning a gazillion git rev-parse --git-path instances, we cache the returned paths, priming the cache upon startup in a single git rev-parse invocation with the known entries.

This fixes https://github.com/git-for-windows/git/issues/1755

Initial-patch-by Philipp Gortan <[email protected]>

Signed-off-by Johannes Schindelin <[email protected]>

---
 git-gui.sh | 52 +++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 49 insertions(+), 3 deletions(-)

git-gui.sh +1 -0

index fd476b6999..8b72e59cd0 100755
@@ -158,6 +158,7 @@ if {[tk windowingsystem] eq "aqua"} {
 
 set _appname {Git Gui}
 set _gitdir {}
+array set _gitdir_cache {}
 set _gitworktree {}
 set _isbare {}
 set _gitexec {}

git-gui.sh +8 -0

+		set gitdir_keys [list \
+			CHERRY_PICK_HEAD FETCH_HEAD GITGUI_BCK GITGUI_EDITMSG \
+			GITGUI_MSG HEAD hooks hooks/prepare-commit-msg \
+			index.lock info info/exclude logs MERGE_HEAD MERGE_MSG \
+			MERGE_RR objects "objects/4\[0-1\]/*" \
+			"objects/4\[0-3\]/*" objects/info \
+			objects/info/alternates objects/pack packed-refs \
+			PREPARE_COMMIT_MSG rebase-merge/head-name remotes \

git-gui.sh +22 -2

+		if {$i eq -1} {
+			set _gitdir $path
+		} else {
+			set _gitdir_cache([lindex $gitdir_keys $i]) $path
+		}
+		incr i
+	}
+}
+
 proc gitdir {args} {
-	global _gitdir
+	global _gitdir _gitdir_cache
+
 	if {$args eq {}} {
 		return $_gitdir
 	}
-	return [eval [list file join $_gitdir] $args]
+
+	set args [eval [list file join] $args]
+	if {![info exists _gitdir_cache($args)]} {
+		if {[package vcompare $::_git_version 2.5.0] >= 0} {
+			set _gitdir_cache($args) [git rev-parse --git-path $args]
+		} else {
+			set _gitdir_cache($args) [file join $_gitdir $args]
+		}
+	}
+
+	return $_gitdir_cache($args)
 }
 
 proc gitexec {args} {

Could you please change the commit subject to more clearly state that we are caching all paths. This is not something just related to hooks any more.

Quoted 17 lines

From: Johannes Schindelin <[email protected]>

Since v2.9.0, Git knows about the config variable core.hookspath that allows overriding the path to the directory containing the Git hooks.

Since v2.10.0, the --git-path option respects that config variable, too, so we may just as well use that command.

For Git versions older than v2.5.0 (which was the first version to support the --git-path option for the rev-parse command), we simply fall back to the previous code.

An original patch handled only the hooksPath setting, however, during the code submission it was deemed better to fix all call to the gitdir function.

Quoted 3 lines

To avoid spawning a gazillion git rev-parse --git-path instances, we cache the returned paths, priming the cache upon startup in a single git rev-parse invocation with the known entries.

I think it would also be a good idea to mention that we are fixing worktree paths not being correct.

Quoted 24 lines

This fixes https://github.com/git-for-windows/git/issues/1755

Initial-patch-by Philipp Gortan <[email protected]>

Signed-off-by Johannes Schindelin <[email protected]>

---
 git-gui.sh | 52 +++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 49 insertions(+), 3 deletions(-)

git-gui.sh +1 -0

index fd476b6999..8b72e59cd0 100755
@@ -158,6 +158,7 @@ if {[tk windowingsystem] eq "aqua"} {
 
 set _appname {Git Gui}
 set _gitdir {}
+array set _gitdir_cache {}
 set _gitworktree {}
 set _isbare {}
 set _gitexec {}

@@ -197,12 +198,57 @@ proc appname {} {
 	return $_appname
 }
 

git-gui.sh proc appname {} { · line 207

Quoted 7 lines

git-gui.sh +7 -0

+proc prime_gitdir_cache {} {
+	global _gitdir _gitdir_cache
+
+	set gitdir_cmd [list git rev-parse --git-dir]
+
+	# `--git-path` is only supported since Git v2.5.0
+	if {[package vcompare $::_git_version 2.5.0] >= 0} {

I think we should mention the source of the list of "magic" keys. A comment mentioning this list came from looking at the common calls to gitdir in the rest of the git-gui code would explain this function to a future reader better.

Quoted 8 lines

git-gui.sh +8 -0

+		set gitdir_keys [list \
+			CHERRY_PICK_HEAD FETCH_HEAD GITGUI_BCK GITGUI_EDITMSG \
+			GITGUI_MSG HEAD hooks hooks/prepare-commit-msg \
+			index.lock info info/exclude logs MERGE_HEAD MERGE_MSG \
+			MERGE_RR objects "objects/4\[0-1\]/*" \
+			"objects/4\[0-3\]/*" objects/info \
+			objects/info/alternates objects/pack packed-refs \
+			PREPARE_COMMIT_MSG rebase-merge/head-name remotes \

git-gui.sh proc appname {} { · line 225

Quoted 10 lines

git-gui.sh +10 -0

+			rr-cache rr-cache/MERGE_RR SQUASH_MSG \
+		]
+
+		foreach key $gitdir_keys {
+			lappend gitdir_cmd --git-path $key
+		}
+	}
+
+	set i -1
+	foreach path [split [eval $gitdir_cmd] "\n"] {

A call to the procedure 'git' is wrapped in a 'catch' in a lot of places. But it is also not wrapped in 'catch' in a lot of other places.

I'm not sure how something like this would fail, so I'm not sure if wrapping this call in a catch is a good idea. But it is something to consider.

Quoted 30 lines

git-gui.sh +22 -2

+		if {$i eq -1} {
+			set _gitdir $path
+		} else {
+			set _gitdir_cache([lindex $gitdir_keys $i]) $path
+		}
+		incr i
+	}
+}
+
 proc gitdir {args} {
-	global _gitdir
+	global _gitdir _gitdir_cache
+
 	if {$args eq {}} {
 		return $_gitdir
 	}
-	return [eval [list file join $_gitdir] $args]
+
+	set args [eval [list file join] $args]
+	if {![info exists _gitdir_cache($args)]} {
+		if {[package vcompare $::_git_version 2.5.0] >= 0} {
+			set _gitdir_cache($args) [git rev-parse --git-path $args]
+		} else {
+			set _gitdir_cache($args) [file join $_gitdir $args]
+		}
+	}
+
+	return $_gitdir_cache($args)
 }
 

git-gui.sh if {[catch { · line 1294

Quoted 10 lines

git-gui.sh +1 -1

 proc gitexec {args} {

@@ -1242,7 +1288,7 @@ if {[catch {
 	&& [catch {
 		# beware that from the .git dir this sets _gitdir to .
 		# and _prefix to the empty string
-		set _gitdir [git rev-parse --git-dir]
+		prime_gitdir_cache
 		set _prefix [git rev-parse --show-prefix]
 	} err]} {
 	load_config 1

Can these paths we cache change while git-gui is running, say by a command run by the user in the terminal? In that case, we should refresh the list when the user rescans.

Other than some minor comments, looks good. Thanks.

--
Regards,
Pratyush Yadav

From: Johannes Schindelin <[email protected]>

Since v2.9.0, Git knows about the config variable core.hookspath that allows overriding the path to the directory containing the Git hooks.

Since v2.10.0, the --git-path option respects that config variable, too, so we may just as well use that command.

Other paths inside .git are equally subject to differ from <gitdir>/<path>, i.e. inside worktrees, where _some_ paths live in the "gitdir" and some live in the "commondir" (i.e. the "gitdir" of the main worktree).

For Git versions older than v2.5.0 (which was the first version to support the --git-path option for the rev-parse command), we simply fall back to the previous code.

An original patch handled only the hooksPath setting, however, during the code submission it was deemed better to fix all call to the gitdir function.

To avoid spawning a gazillion git rev-parse --git-path instances, we cache the returned paths, priming the cache upon startup in a single git rev-parse invocation with some paths (that have been determined via a typical startup and via grepping the source code for calls to the gitdir function).

This fixes https://github.com/git-for-windows/git/issues/1755

Initial-patch-by Philipp Gortan <[email protected]>

Signed-off-by Johannes Schindelin <[email protected]>

---
 git-gui.sh | 62 ++++++++++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 58 insertions(+), 4 deletions(-)

git-gui.sh +50 -2

index fd476b6999..c684dc7ae1 100755
@@ -158,6 +158,7 @@ if {[tk windowingsystem] eq "aqua"} {
 
 set _appname {Git Gui}
 set _gitdir {}
+array set _gitdir_cache {}
 set _gitworktree {}
 set _isbare {}
 set _gitexec {}

@@ -197,12 +198,59 @@ proc appname {} {
 	return $_appname
 }
 
+proc prime_gitdir_cache {} {
+	global _gitdir _gitdir_cache
+
+	set gitdir_cmd [list git rev-parse --git-dir]
+
+	# `--git-path` is only supported since Git v2.5.0
+	if {[package vcompare $::_git_version 2.5.0] >= 0} {
+		# This list was generated from a typical startup as well as from
+		# grepping through Git GUI's source code.
+		set gitdir_keys [list \
+			CHERRY_PICK_HEAD FETCH_HEAD GITGUI_BCK GITGUI_EDITMSG \
+			GITGUI_MSG HEAD hooks hooks/prepare-commit-msg \
+			index.lock info info/exclude logs MERGE_HEAD MERGE_MSG \
+			MERGE_RR objects "objects/4\[0-1\]/*" \
+			"objects/4\[0-3\]/*" objects/info \
+			objects/info/alternates objects/pack packed-refs \
+			PREPARE_COMMIT_MSG rebase-merge/head-name remotes \
+			rr-cache rr-cache/MERGE_RR SQUASH_MSG \
+		]
+
+		foreach key $gitdir_keys {
+			lappend gitdir_cmd --git-path $key
+		}
+	}
+
+	set i -1
+	foreach path [split [eval $gitdir_cmd] "\n"] {
+		if {$i eq -1} {
+			set _gitdir $path
+		} else {
+			set _gitdir_cache([lindex $gitdir_keys $i]) $path
+		}
+		incr i
+	}
+}
+
 proc gitdir {args} {
-	global _gitdir
+	global _gitdir _gitdir_cache
+
 	if {$args eq {}} {
 		return $_gitdir
 	}
-	return [eval [list file join $_gitdir] $args]
+
+	set args [eval [list file join] $args]
+	if {![info exists _gitdir_cache($args)]} {
+		if {[package vcompare $::_git_version 2.5.0] >= 0} {
+			set _gitdir_cache($args) [git rev-parse --git-path $args]
+		} else {
+			set _gitdir_cache($args) [file join $_gitdir $args]
+		}
+	}
+
+	return $_gitdir_cache($args)
 }
 
 proc gitexec {args} {

git-gui.sh +4 -0

+		array unset _gitdir_cache
+		prime_gitdir_cache
+	}
+
 	repository_state newType newHEAD newMERGE_HEAD
 	if {[string match amend* $commit_type]
 		&& $newType eq {normal}

git-gui.sh proc rescan {after {honor_trustmtime 1}} { · line 1509

Quoted 11 lines · Johannes Schindelin via GitGitGadget

git-gui.sh +3 -1

@@ -1453,10 +1501,16 @@ proc rescan {after {honor_trustmtime 1}} {
 	global HEAD PARENT MERGE_HEAD commit_type
 	global ui_index ui_workdir ui_comm
 	global rescan_active file_states
-	global repo_config
+	global repo_config _gitdir_cache
 
 	if {$rescan_active > 0 || ![lock_index read]} return
 
+	# Only re-prime gitdir cache on a full rescan
+	if {$after ne "ui_ready"} {

What do you mean by a "full rescan"? I assume you use it as the differentiator between ui_do_rescan (called when you hit F5 or choose rescan from the menu) and do_rescan (called when you revert a line or hunk), and a "full rescan" refers to ui_do_rescan.

Well in that case, this check is incorrect. do_rescan passes only "ui_ready" and ui_do_rescan passes "force_first_diff ui_ready".

But either way, I'm not a big fan of this. This check makes assumptions about the behaviour of its callers based on what they pass to $after. The way I see it, $after should be a black box to rescan, and it should make absolutely no assumptions about it.

Doing it this way is really brittle, and would break as soon as someone changes the behaviour of ui_do_rescan. If someone in the future passes a different value in $after, this would stop working as intended and would not refresh the cached list on a rescan.

So, I think a better place for this if statement would be in ui_do_rescan. This would mean adding a new function that does this. But if we unset _gitdir_cache in prime_gitdir_cache (I see no reason not to), we can get away with just something like:

  proc ui_do_rescan {} {
  	rescan {prime_gitdir_cache; ui_ready}
  }

Though since prime_gitdir_cache does not really depend on the rescan being finished, something like this would also work fine:

  proc ui_do_rescan {} {
  	rescan ui_ready
  	prime_gitdir_cache
  }

This would allow us to do these two things in parallel since rescan is asynchronous. But that would also mean it is possible that the status bar would show "Ready" while prime_gitdir_cache is still executing.

I can't really make up my mind on what is better. I'm inclining on using the latter way, effectively trading a bit of UI inconsistency for performance (at least in theory).

Thoughts?

git-gui.sh proc rescan {after {honor_trustmtime 1}} { · line 1516

Quoted 7 lines

git-gui.sh +4 -0

+		array unset _gitdir_cache
+		prime_gitdir_cache
+	}
+
 	repository_state newType newHEAD newMERGE_HEAD
 	if {[string match amend* $commit_type]
 		&& $newType eq {normal}

--
Regards,
Pratyush Yadav

Hi Pratyush,

Quoted 41 lines · Pratyush Yadav

git-gui.sh proc rescan {after {honor_trustmtime 1}} { · line 1509

Quoted 11 lines · Johannes Schindelin via GitGitGadget

git-gui.sh +3 -1

@@ -1453,10 +1501,16 @@ proc rescan {after {honor_trustmtime 1}} {
 	global HEAD PARENT MERGE_HEAD commit_type
 	global ui_index ui_workdir ui_comm
 	global rescan_active file_states
-	global repo_config
+	global repo_config _gitdir_cache
 
 	if {$rescan_active > 0 || ![lock_index read]} return
 
+	# Only re-prime gitdir cache on a full rescan
+	if {$after ne "ui_ready"} {

What do you mean by a "full rescan"? I assume you use it as the differentiator between ui_do_rescan (called when you hit F5 or choose rescan from the menu) and do_rescan (called when you revert a line or hunk), and a "full rescan" refers to ui_do_rescan.

Well in that case, this check is incorrect. do_rescan passes only "ui_ready" and ui_do_rescan passes "force_first_diff ui_ready".

But either way, I'm not a big fan of this. This check makes assumptions about the behaviour of its callers based on what they pass to $after. The way I see it, $after should be a black box to rescan, and it should make absolutely no assumptions about it.

Doing it this way is really brittle, and would break as soon as someone changes the behaviour of ui_do_rescan. If someone in the future passes a different value in $after, this would stop working as intended and would not refresh the cached list on a rescan.

So, I think a better place for this if statement would be in ui_do_rescan. This would mean adding a new function that does this. But if we unset _gitdir_cache in prime_gitdir_cache (I see no reason not to), we can get away with just something like:

  proc ui_do_rescan {} {
  	rescan {prime_gitdir_cache; ui_ready}
  }

Quoted 7 lines

Though since prime_gitdir_cache does not really depend on the rescan being finished, something like this would also work fine:

  proc ui_do_rescan {} {
  	rescan ui_ready
  	prime_gitdir_cache
  }

That was my first attempt. However, there is a very important piece of code that is even still quoted above: that if {$rescan_active > 0 || ![lock_index read]} return part.

I do _not_ want to interfere with an actively-going-on rescan. If there is an active one, I don't want to re-prime the _gitdir cache.

That was the reason why I put the additional code into rescan rather than into ui_do_rescan().

Ciao,
Johannes

Quoted 18 lines

This would allow us to do these two things in parallel since rescan is asynchronous. But that would also mean it is possible that the status bar would show "Ready" while prime_gitdir_cache is still executing.

I can't really make up my mind on what is better. I'm inclining on using the latter way, effectively trading a bit of UI inconsistency for performance (at least in theory).

Thoughts?

Quoted 7 lines

git-gui.sh +4 -0

+		array unset _gitdir_cache
+		prime_gitdir_cache
+	}
+
 	repository_state newType newHEAD newMERGE_HEAD
 	if {[string match amend* $commit_type]
 		&& $newType eq {normal}

Quoted 57 lines · Johannes Schindelin

Hi Pratyush,

Quoted 41 lines · Pratyush Yadav

git-gui.sh proc rescan {after {honor_trustmtime 1}} { · line 1509

Quoted 11 lines · Johannes Schindelin via GitGitGadget

git-gui.sh +3 -1

@@ -1453,10 +1501,16 @@ proc rescan {after {honor_trustmtime 1}} {
 	global HEAD PARENT MERGE_HEAD commit_type
 	global ui_index ui_workdir ui_comm
 	global rescan_active file_states
-	global repo_config
+	global repo_config _gitdir_cache
 
 	if {$rescan_active > 0 || ![lock_index read]} return
 
+	# Only re-prime gitdir cache on a full rescan
+	if {$after ne "ui_ready"} {

What do you mean by a "full rescan"? I assume you use it as the differentiator between ui_do_rescan (called when you hit F5 or choose rescan from the menu) and do_rescan (called when you revert a line or hunk), and a "full rescan" refers to ui_do_rescan.

Well in that case, this check is incorrect. do_rescan passes only "ui_ready" and ui_do_rescan passes "force_first_diff ui_ready".

But either way, I'm not a big fan of this. This check makes assumptions about the behaviour of its callers based on what they pass to $after. The way I see it, $after should be a black box to rescan, and it should make absolutely no assumptions about it.

Doing it this way is really brittle, and would break as soon as someone changes the behaviour of ui_do_rescan. If someone in the future passes a different value in $after, this would stop working as intended and would not refresh the cached list on a rescan.

So, I think a better place for this if statement would be in ui_do_rescan. This would mean adding a new function that does this. But if we unset _gitdir_cache in prime_gitdir_cache (I see no reason not to), we can get away with just something like:

  proc ui_do_rescan {} {
  	rescan {prime_gitdir_cache; ui_ready}
  }

Quoted 7 lines

Though since prime_gitdir_cache does not really depend on the rescan being finished, something like this would also work fine:

  proc ui_do_rescan {} {
  	rescan ui_ready
  	prime_gitdir_cache
  }

That was my first attempt. However, there is a very important piece of code that is even still quoted above: that if {$rescan_active > 0 || ![lock_index read]} return part.

Quoted 2 lines

I do _not_ want to interfere with an actively-going-on rescan. If there is an active one, I don't want to re-prime the _gitdir cache.

Good catch! In that case I suppose refreshing the cache in $after would be the way to go (IOW, the former of my two suggestions). Anything $after won't get executed if we return early from that check.

Quoted 18 lines

That was the reason why I put the additional code into rescan rather than into ui_do_rescan().

Ciao,
Johannes

Quoted 11 lines

This would allow us to do these two things in parallel since rescan is asynchronous. But that would also mean it is possible that the status bar would show "Ready" while prime_gitdir_cache is still executing.

I can't really make up my mind on what is better. I'm inclining on using the latter way, effectively trading a bit of UI inconsistency for performance (at least in theory).

Thoughts?

Quoted 7 lines

Quoted 7 lines

git-gui.sh proc rescan {after {honor_trustmtime 1}} { · line 1516

Quoted 7 lines

git-gui.sh +4 -0

+		array unset _gitdir_cache
+		prime_gitdir_cache
+	}
+
 	repository_state newType newHEAD newMERGE_HEAD
 	if {[string match amend* $commit_type]
 		&& $newType eq {normal}

--
Regards,
Pratyush Yadav

Quoted 56 lines · Johannes Schindelin

Quoted 41 lines · Pratyush Yadav

git-gui.sh proc rescan {after {honor_trustmtime 1}} { · line 1509

Quoted 11 lines · Johannes Schindelin via GitGitGadget

git-gui.sh +3 -1

@@ -1453,10 +1501,16 @@ proc rescan {after {honor_trustmtime 1}} {
 	global HEAD PARENT MERGE_HEAD commit_type
 	global ui_index ui_workdir ui_comm
 	global rescan_active file_states
-	global repo_config
+	global repo_config _gitdir_cache
 
 	if {$rescan_active > 0 || ![lock_index read]} return
 
+	# Only re-prime gitdir cache on a full rescan
+	if {$after ne "ui_ready"} {

What do you mean by a "full rescan"? I assume you use it as the differentiator between ui_do_rescan (called when you hit F5 or choose rescan from the menu) and do_rescan (called when you revert a line or hunk), and a "full rescan" refers to ui_do_rescan.

Well in that case, this check is incorrect. do_rescan passes only "ui_ready" and ui_do_rescan passes "force_first_diff ui_ready".

But either way, I'm not a big fan of this. This check makes assumptions about the behaviour of its callers based on what they pass to $after. The way I see it, $after should be a black box to rescan, and it should make absolutely no assumptions about it.

Doing it this way is really brittle, and would break as soon as someone changes the behaviour of ui_do_rescan. If someone in the future passes a different value in $after, this would stop working as intended and would not refresh the cached list on a rescan.

So, I think a better place for this if statement would be in ui_do_rescan. This would mean adding a new function that does this. But if we unset _gitdir_cache in prime_gitdir_cache (I see no reason not to), we can get away with just something like:

  proc ui_do_rescan {} {
  	rescan {prime_gitdir_cache; ui_ready}
  }

Quoted 7 lines

Though since prime_gitdir_cache does not really depend on the rescan being finished, something like this would also work fine:

  proc ui_do_rescan {} {
  	rescan ui_ready
  	prime_gitdir_cache
  }

That was my first attempt. However, there is a very important piece of code that is even still quoted above: that if {$rescan_active > 0 || ![lock_index read]} return part.

Quoted 2 lines

I do _not_ want to interfere with an actively-going-on rescan. If there is an active one, I don't want to re-prime the _gitdir cache.

Good catch! In that case I suppose refreshing the cache in $after would be the way to go (IOW, the former of my two suggestions). Anything $after won't get executed if we return early from that check.

Quoted 18 lines

That was the reason why I put the additional code into rescan rather than into ui_do_rescan().

Ciao,
Johannes

Quoted 11 lines

This would allow us to do these two things in parallel since rescan is asynchronous. But that would also mean it is possible that the status bar would show "Ready" while prime_gitdir_cache is still executing.

I can't really make up my mind on what is better. I'm inclining on using the latter way, effectively trading a bit of UI inconsistency for performance (at least in theory).

Thoughts?

Quoted 7 lines

Quoted 7 lines

git-gui.sh proc rescan {after {honor_trustmtime 1}} { · line 1516

Quoted 7 lines

git-gui.sh +4 -0

+		array unset _gitdir_cache
+		prime_gitdir_cache
+	}
+
 	repository_state newType newHEAD newMERGE_HEAD
 	if {[string match amend* $commit_type]
 		&& $newType eq {normal}

--
Regards,
Pratyush Yadav

Hi Pratyush,

Quoted 58 lines · Pratyush Yadav

Quoted 57 lines · Johannes Schindelin

Hi Pratyush,

Quoted 41 lines · Pratyush Yadav

git-gui.sh proc rescan {after {honor_trustmtime 1}} { · line 1509

Quoted 11 lines · Johannes Schindelin via GitGitGadget

git-gui.sh +3 -1

@@ -1453,10 +1501,16 @@ proc rescan {after {honor_trustmtime 1}} {
 	global HEAD PARENT MERGE_HEAD commit_type
 	global ui_index ui_workdir ui_comm
 	global rescan_active file_states
-	global repo_config
+	global repo_config _gitdir_cache
 
 	if {$rescan_active > 0 || ![lock_index read]} return
 
+	# Only re-prime gitdir cache on a full rescan
+	if {$after ne "ui_ready"} {

What do you mean by a "full rescan"? I assume you use it as the differentiator between ui_do_rescan (called when you hit F5 or choose rescan from the menu) and do_rescan (called when you revert a line or hunk), and a "full rescan" refers to ui_do_rescan.

Well in that case, this check is incorrect. do_rescan passes only "ui_ready" and ui_do_rescan passes "force_first_diff ui_ready".

But either way, I'm not a big fan of this. This check makes assumptions about the behaviour of its callers based on what they pass to $after. The way I see it, $after should be a black box to rescan, and it should make absolutely no assumptions about it.

Doing it this way is really brittle, and would break as soon as someone changes the behaviour of ui_do_rescan. If someone in the future passes a different value in $after, this would stop working as intended and would not refresh the cached list on a rescan.

So, I think a better place for this if statement would be in ui_do_rescan. This would mean adding a new function that does this. But if we unset _gitdir_cache in prime_gitdir_cache (I see no reason not to), we can get away with just something like:

  proc ui_do_rescan {} {
  	rescan {prime_gitdir_cache; ui_ready}
  }

Quoted 7 lines

Though since prime_gitdir_cache does not really depend on the rescan being finished, something like this would also work fine:

  proc ui_do_rescan {} {
  	rescan ui_ready
  	prime_gitdir_cache
  }

That was my first attempt. However, there is a very important piece of code that is even still quoted above: that if {$rescan_active > 0 || ![lock_index read]} return part.

Quoted 6 lines

Quoted 2 lines

I do _not_ want to interfere with an actively-going-on rescan. If there is an active one, I don't want to re-prime the _gitdir cache.

Good catch! In that case I suppose refreshing the cache in $after would be the way to go (IOW, the former of my two suggestions). Anything $after won't get executed if we return early from that check.

The obvious problem with that solution is that the _gitdir is reset _after_ the rescan. In my solution, it is reset _before_, as I have no idea how often the _gitdir values are used during a rescan (and if none of they were, I would like to be very cautious to prepare for a future where any of those _gitdir values _are_ used during a rescan).

So I am afraid that I find way too serious problems with both of your proposed alternatives.