Re: On refreshing the index

From: Sergio Callegari <sergio.callegari@gmail.com>
Date: Mar 15, 2010, 00:19 UTC
Message-ID: <4B9D7CF5.5010404@gmail.com>
In-Reply-To: <7vmxyb3la7.fsf@alter.siamese.dyndns.org>

Junio C Hamano wrote:

Show 28 quoted lines

> Sergio Callegari <sergio.callegari@gmail.com> writes:
>
>   
>> If I run git status, git runs filters on a couple of opendocument files for
>> which a filter is defined
>>
>> GIT_TRACE=1 git status
>> trace: built-in: git 'status'
>> trace: run_command: 'rezip -p ODF_UNCOMPRESS2'
>> trace: exec: 'sh' '-c' 'rezip -p ODF_UNCOMPRESS2' 'rezip -p ODF_UNCOMPRESS2'
>> trace: run_command: 'rezip -p ODF_UNCOMPRESS2'
>> trace: exec: 'sh' '-c' 'rezip -p ODF_UNCOMPRESS2' 'rezip -p ODF_UNCOMPRESS2'
>> # On branch M05
>> # Untracked files:
>> #   (use "git add <file>..." to include in what will be committed)
>> #
>> #       WIP/
>> #       program.txt
>> #       program.txt~
>> nothing added to commit but untracked files present (use "git add" to track)
>>     
>
> What does "git diff-files" and/or "git diff-index HEAD" say at this point?
> If they do not say there are no difference, that means that the file on
> the filesystem and the blob registered in the index are different, even
> though after transmogrified with rezip (whatever it does) these two
> different blobs may look the same.
>

Neither git diff-files" and/or "git diff-index HEAD say nothing at this point...

> I think the difference between "may look the same" and "identical" is what
> you are seeing.  Try "git add" on those paths and see what happens.
>

Is there any way I can find out which file is the guilty one since git diff-files says nothing? E.g. a trace telling me on what is the filter being called?

BTW... some notes that may be useful...

1) rezip is a mere recompressor.  It takes a zip file and re-creates it 
at zero compression, so that the git delta logic can do a good job on it 
on repacking.  I've found this useful on zip files, openoffice files, 
jar files, etc.

2) if I clone outside git, git update-index --refresh is always ok at 
making git status fast (i.e. not running expensive filters).

3) The problem always happens when I switch branches, right after the 
switch.

Sergio