Re: git gui error with relocated repository

Chris Packham [off-list ref] writes:

On Tue, Apr 29, 2014 at 2:56 PM, Chris Packham [off-list ref] wrote:

quoted

Hi Pat,

I'm running git 2.0.0-rc0 (haven't got round to pulling down rc1 yet)
which includes gitgui-0.19.0 and I'm getting a new error when I run
'git gui' in a repository with a .git file (created by git submodule).

I can send you a screencap of the error message off list if you want
but the text is

"No working directory ../../../<repo>

couldn't change working directory to ../../../<repo>: no such file or directory"

My tcl is a little rusty but I think the problem might be this snippet.

# v1.7.0 introduced --show-toplevel to return the canonical work-tree
if {[package vsatisfies $_git_version 1.7.0]} {
   if { [is_Cygwin] } {
       catch {set _gitworktree [exec cygpath --windows [git rev-parse
--show-toplevel]]}
   } else {
       set _gitworktree [git rev-parse --show-toplevel]
   }
} else {
   # try to set work tree from environment, core.worktree or use
   # cdup to obtain a relative path to the top of the worktree. If
   # run from the top, the ./ prefix ensures normalize expands pwd.
   if {[catch { set _gitworktree $env(GIT_WORK_TREE) }]} {
       set _gitworktree [get_config core.worktree]
       if {$_gitworktree eq ""} {
           set _gitworktree [file normalize ./[git rev-parse --show-cdup]]
       }
   }
}

The  vsatisfies call probably doesn't handle '2.0.0.rc0' and the
fallback behaviour probably needs to normalise core.worktree

The _git_version variable has already been trimmed to remove such
suffixes so the version comparison here will be ok. It looks more likely
to be something to do with the .git being a file with a link being
mishandled. How did you setup this test repository with its link to a
parent?

quoted

Here's some other info that might help

  $ git --version
  git version 2.0.0.rc0

  $ cat .git
  gitdir: ../.git/modules/<repo>

  $ git rev-parse --git-dir
  /home/chris/src/<super>/.git/modules/<repo>

  $ git config core.worktree
  ../../../<repo>

Thanks,
Chris

-- 
Pat Thoyts                            http://www.patthoyts.tk/
PGP fingerprint 2C 6E 98 07 2C 59 C8 97  10 CE 11 E6 04 E0 B9 DD

On 07/05/14 00:10, Pat Thoyts wrote:

Chris Packham [off-list ref] writes:

quoted

On Tue, Apr 29, 2014 at 2:56 PM, Chris Packham [off-list ref] wrote:

quoted

Hi Pat,

I'm running git 2.0.0-rc0 (haven't got round to pulling down rc1 yet)
which includes gitgui-0.19.0 and I'm getting a new error when I run
'git gui' in a repository with a .git file (created by git submodule).

I can send you a screencap of the error message off list if you want
but the text is

"No working directory ../../../<repo>

couldn't change working directory to ../../../<repo>: no such file or directory"

My tcl is a little rusty but I think the problem might be this snippet.

# v1.7.0 introduced --show-toplevel to return the canonical work-tree
if {[package vsatisfies $_git_version 1.7.0]} {
   if { [is_Cygwin] } {
       catch {set _gitworktree [exec cygpath --windows [git rev-parse
--show-toplevel]]}
   } else {
       set _gitworktree [git rev-parse --show-toplevel]
   }
} else {
   # try to set work tree from environment, core.worktree or use
   # cdup to obtain a relative path to the top of the worktree. If
   # run from the top, the ./ prefix ensures normalize expands pwd.
   if {[catch { set _gitworktree $env(GIT_WORK_TREE) }]} {
       set _gitworktree [get_config core.worktree]
       if {$_gitworktree eq ""} {
           set _gitworktree [file normalize ./[git rev-parse --show-cdup]]
       }
   }
}

The  vsatisfies call probably doesn't handle '2.0.0.rc0' and the
fallback behaviour probably needs to normalise core.worktree

The _git_version variable has already been trimmed to remove such
suffixes so the version comparison here will be ok. 

I don't think that's true 'git rev-parse --show-toplevel' does the right
thing - if it's run.

It looks more likely
to be something to do with the .git being a file with a link being
mishandled. How did you setup this test repository with its link to a
parent?

It's just a git clone of the parent and a git submodule init. A subtle
thing to notice is that config.worktree is relative to $GIT_DIR. All the
configuration was done by git without any intervention from me.

quoted

quoted

Here's some other info that might help

  $ git --version
  git version 2.0.0.rc0

  $ cat .git
  gitdir: ../.git/modules/<repo>

  $ git rev-parse --git-dir
  /home/chris/src/<super>/.git/modules/<repo>

  $ git config core.worktree
  ../../../<repo>

Thanks,
Chris

On 07/05/14 19:28, Chris Packham wrote:

On 07/05/14 00:10, Pat Thoyts wrote:

quoted

Chris Packham [off-list ref] writes:

quoted

On Tue, Apr 29, 2014 at 2:56 PM, Chris Packham [off-list ref] wrote:

quoted

Hi Pat,

I'm running git 2.0.0-rc0 (haven't got round to pulling down rc1 yet)
which includes gitgui-0.19.0 and I'm getting a new error when I run
'git gui' in a repository with a .git file (created by git submodule).

I can send you a screencap of the error message off list if you want
but the text is

"No working directory ../../../<repo>

couldn't change working directory to ../../../<repo>: no such file or directory"

My tcl is a little rusty but I think the problem might be this snippet.

# v1.7.0 introduced --show-toplevel to return the canonical work-tree
if {[package vsatisfies $_git_version 1.7.0]} {
   if { [is_Cygwin] } {
       catch {set _gitworktree [exec cygpath --windows [git rev-parse
--show-toplevel]]}
   } else {
       set _gitworktree [git rev-parse --show-toplevel]
   }
} else {
   # try to set work tree from environment, core.worktree or use
   # cdup to obtain a relative path to the top of the worktree. If
   # run from the top, the ./ prefix ensures normalize expands pwd.
   if {[catch { set _gitworktree $env(GIT_WORK_TREE) }]} {
       set _gitworktree [get_config core.worktree]
       if {$_gitworktree eq ""} {
           set _gitworktree [file normalize ./[git rev-parse --show-cdup]]
       }
   }
}

The  vsatisfies call probably doesn't handle '2.0.0.rc0' and the
fallback behaviour probably needs to normalise core.worktree

The _git_version variable has already been trimmed to remove such
suffixes so the version comparison here will be ok. 

I don't think that's true 'git rev-parse --show-toplevel' does the right
thing - if it's run.

We'll the trimming works but vstatisfies doesn't

  puts $_git_version
  puts [package vsatisfies $_git_version 1.7.0]

  2.0.0
  0

quoted

It looks more likely
to be something to do with the .git being a file with a link being
mishandled. How did you setup this test repository with its link to a
parent?

It's just a git clone of the parent and a git submodule init. A subtle
thing to notice is that config.worktree is relative to $GIT_DIR. All the
configuration was done by git without any intervention from me.

quoted

quoted

quoted

Here's some other info that might help

  $ git --version
  git version 2.0.0.rc0

  $ cat .git
  gitdir: ../.git/modules/<repo>

  $ git rev-parse --git-dir
  /home/chris/src/<super>/.git/modules/<repo>

  $ git config core.worktree
  ../../../<repo>

Thanks,
Chris

Since git 2.0.0 starting git gui in a submodule using a gitfile fails with
the following error:

   No working directory ../../../<path>

   couldn't change working directory
   to "../../../<path>": no such file or
   directory

This is because "git rev-parse --show-toplevel" is only run when git gui
sees a git version of at least 1.7.0 (which is the version in which the
--show-toplevel option was introduced). But it uses vsatisfies to check
that, which is documented to return false when the major version changes,
which is not what we want here.

Change vsatisfies to vcompare when testing for a git version to avoid the
problem when the major version changes (but still use vsatisfies in those
places where the Tk version is checked). This is done for both the place
that caused the reported bug and another spot where the git version is
tested for another feature.

Reported-by: Chris Packham <redacted>
Reported-by: Yann Dirson <redacted>
Signed-off-by: Jens Lehmann <redacted>
---

Am 07.05.2014 09:49, schrieb Chris Packham:

On 07/05/14 19:28, Chris Packham wrote:

quoted

On 07/05/14 00:10, Pat Thoyts wrote:

quoted

Chris Packham [off-list ref] writes:

quoted

On Tue, Apr 29, 2014 at 2:56 PM, Chris Packham [off-list ref] wrote:

quoted

Hi Pat,

I'm running git 2.0.0-rc0 (haven't got round to pulling down rc1 yet)
which includes gitgui-0.19.0 and I'm getting a new error when I run
'git gui' in a repository with a .git file (created by git submodule).

I can send you a screencap of the error message off list if you want
but the text is

"No working directory ../../../<repo>

couldn't change working directory to ../../../<repo>: no such file or directory"

My tcl is a little rusty but I think the problem might be this snippet.

# v1.7.0 introduced --show-toplevel to return the canonical work-tree
if {[package vsatisfies $_git_version 1.7.0]} {
   if { [is_Cygwin] } {
       catch {set _gitworktree [exec cygpath --windows [git rev-parse
--show-toplevel]]}
   } else {
       set _gitworktree [git rev-parse --show-toplevel]
   }
} else {
   # try to set work tree from environment, core.worktree or use
   # cdup to obtain a relative path to the top of the worktree. If
   # run from the top, the ./ prefix ensures normalize expands pwd.
   if {[catch { set _gitworktree $env(GIT_WORK_TREE) }]} {
       set _gitworktree [get_config core.worktree]
       if {$_gitworktree eq ""} {
           set _gitworktree [file normalize ./[git rev-parse --show-cdup]]
       }
   }
}

The  vsatisfies call probably doesn't handle '2.0.0.rc0' and the
fallback behaviour probably needs to normalise core.worktree

The _git_version variable has already been trimmed to remove such
suffixes so the version comparison here will be ok. 

I don't think that's true 'git rev-parse --show-toplevel' does the right
thing - if it's run.

We'll the trimming works but vstatisfies doesn't

  puts $_git_version
  puts [package vsatisfies $_git_version 1.7.0]

  2.0.0
  0

Yup, looks like vsatisfies is doing just what it is supposed to (according
to http://www.astro.princeton.edu/~rhl/Tcl-Tk_docs/tcl/package.n.html):

   package vsatisfies version1 version2
       Returns 1 if scripts written for version2 will work unchanged
       with version1 (i.e. version1 is equal to or greater than version2
       and they both have the same major version number), 0 otherwise.

The bump in the major number from 1 to 2 makes vsatisfies assume that the
version is not compatible anymore, I believe we should have used vcompare
here and in another place.


 git-gui.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/git-gui.sh b/git-gui.sh
index cf2209b..ed2418b 100755
--- a/git-gui.sh
+++ b/git-gui.sh

@@ -1283,7 +1283,7 @@ load_config 0
 apply_config

 # v1.7.0 introduced --show-toplevel to return the canonical work-tree
-if {[package vsatisfies $_git_version 1.7.0]} {
+if {[package vcompare $_git_version 1.7.0]} {
 	if { [is_Cygwin] } {
 		catch {set _gitworktree [exec cygpath --windows [git rev-parse --show-toplevel]]}
 	} else {

@@ -1539,7 +1539,7 @@ proc rescan_stage2 {fd after} {
 		close $fd
 	}

-	if {[package vsatisfies $::_git_version 1.6.3]} {
+	if {[package vcompare $::_git_version 1.6.3]} {
 		set ls_others [list --exclude-standard]
 	} else {
 		set ls_others [list --exclude-per-directory=.gitignore]

-- 
2.0.0.rc3.2.g998f840

On 14/05/14 09:24, Jens Lehmann wrote:

Since git 2.0.0 starting git gui in a submodule using a gitfile fails with
the following error:

   No working directory ../../../<path>

   couldn't change working directory
   to "../../../<path>": no such file or
   directory

This is because "git rev-parse --show-toplevel" is only run when git gui
sees a git version of at least 1.7.0 (which is the version in which the
--show-toplevel option was introduced). But it uses vsatisfies to check
that, which is documented to return false when the major version changes,
which is not what we want here.

Change vsatisfies to vcompare when testing for a git version to avoid the
problem when the major version changes (but still use vsatisfies in those
places where the Tk version is checked). This is done for both the place
that caused the reported bug and another spot where the git version is
tested for another feature.

Reported-by: Chris Packham <redacted>
Reported-by: Yann Dirson <redacted>
Signed-off-by: Jens Lehmann <redacted>
---

Looks good to me (refer to previous comment about being rusty with tcl).
More importantly

 Tested-by: Chris Packham [off-list ref]

There is the alternative I posted which just does away with the version
checks (actually that only addressed one of the vcompare call sites).

There are also these checks that I guess are correct (if not
inconsistent) but again who is still using git 1.5.3?

git-gui.sh:1098:        >= 1.5.3 {
lib/blame.tcl:800:                      if {[git-version >= 1.5.3]} {
lib/blame.tcl:849:      if {[git-version >= 1.5.3]} {
lib/checkout_op.tcl:513:        >= 1.5.3 {

Junio, I believe this issue needs to be fixed before 2.0 final. Otherwise
git gui will not work inside submodules anymore due to the major version
number change from 1 to 2. I'd like to hear Pat's opinion on this; even
though I think my patch is less risky (as it doesn't change behavior for
pre-2 versions), he might like Chris' proposal better.

Am 13.05.2014 23:24, schrieb Jens Lehmann: