git archive --worktree-attributes doesn't exclude .gitattributes anymore

> Da: Gianfranco Costamagna <costamagnagianfranco@yahoo.it>
> A: "git@vger.kernel.org" <git@vger.kernel.org>
> Cc: 
> Inviato: Mercoledì 5 Giugno 2013 0:18
> Oggetto: git archive --worktree-attributes doesn't exclude .gitattributes anymore
> 
>g it version 1.8.1.2
> (please cc me, I'm not subscribed to this list)
> 
> 
> Hi Developers, I write here because since my ubuntu update (quantal to raring)
> and git update from 1.7.10.4-1ubuntu1 to 1.8.1.2
> 
> my export script doesn't work anymore.
> 
> I tried to put .gitattributes or .git/info/attributes, the file is the following
> http://pastebin.com/irngA1L8
> 
> the git is
> git clone http://boinc.berkeley.edu/git/boinc-v2.git
> 
> and the command is
> 
> git archive --prefix boinc-7.1.7+dfsg/ --format tgz -o 
> ../boinc_7.1.7+dfsg.orig.tar.gz -9  client_release/7.1/7.1.7
> 
> The archive gets created, but every file is inside, no exclusions at all.
> 
> I suspect a regression between git 1.7 and 1.8
> 
> thanks for your time
> 
> 
> Gianfranco
> 

> git version 1.8.1.2
> [...]
> I tried to put .gitattributes or .git/info/attributes, the file is the following
> http://pastebin.com/irngA1L8
> [...]
> The archive gets created, but every file is inside, no exclusions at all.
> 
> I suspect a regression between git 1.7 and 1.8

>________________________________
> Da: Jeff King <peff@peff.net>
>A: Gianfranco Costamagna <costamagnagianfranco@yahoo.it> 
>Cc: "git@vger.kernel.org" <git@vger.kernel.org> 
>Inviato: Mercoledì 5 Giugno 2013 9:10
>Oggetto: Re: git archive --worktree-attributes doesn't exclude .gitattributes anymore
> 
>
>On Wed, Jun 05, 2013 at 07:10:01AM +0100, Gianfranco Costamagna wrote:
>
>> Just a question, is it possible to add a new test for handling this
>> kind of regressions?
>
>Yes, I added one in commit efa5f825.
>
>> And second, is it possible to have a patch for this problem? I'll be
>> glad to backport in debian, or to take care of updating the whole git
>> package (it is already in experimental BTW)
>
>The fix took a fair bit of refactoring; it's in commits
>9db9eec..efa5f825. You can either cherry-pick that range, or just use
>"git diff" to generate a patch.
>
>However, for Debian, I don't see much point. Everything up through
>unstable does not have the bug (it is all 1.7.10.x or older), and the
>version in experimental is already 1.8.3. So as far as I can tell, no
>Debian release is currently shipping a buggy version.
>
>-Peff
>
>
> 

>Da: Gianfranco Costamagna <costamagnagianfranco@yahoo.it>
>A: "peff@peff.net" <peff@peff.net>; "git@vger.kernel.org" <git@vger.kernel.org> 
>Inviato: Mercoledì 5 Giugno 2013 8:10
>Oggetto: Re: git archive --worktree-attributes doesn't exclude .gitattributes  anymore
> 
>
>
>Oh many thanks for your answer Jeff!
>I was wondering about a fault in my side!
>I was already building 1.8.3 when I sent the first mail...
>Anyway your software has a bunch of tests that make compilation a little slow (and no, I don't want to disable them, they are there for a really good reason).
>Just a question, is it possible to add a new test for handling this kind of regressions?
>And second, is it possible to have a patch for this problem? I'll be glad to backport in debian, or to take care of updating the whole git package (it is already in experimental BTW)
>Thanks
>G.
>Sent from Yahoo! Mail on Android 
>
>
>
>________________________________
> From:  Jeff King <peff@peff.net>; 
>To:  Gianfranco Costamagna <costamagnagianfranco@yahoo.it>; 
>Cc:  git@vger.kernel.org <git@vger.kernel.org>; 
>Subject:  Re: git archive --worktree-attributes doesn't exclude .gitattributes anymore 
>Sent:  Wed, Jun 5, 2013 3:13:49 AM 
>
>
>On Tue, Jun 04, 2013 at 11:18:59PM +0100, Gianfranco Costamagna wrote:
>
>> git version 1.8.1.2
>> [...]
>> I tried to put .gitattributes or .git/info/attributes, the file is the following
>> http://pastebin.com/irngA1L8
>> [...]
>> The archive gets created, but every file is inside, no exclusions at all.
>> 
>> I suspect a regression between git 1.7 and 1.8
>
>Yes, there was a regression with matching directory names in
>gitattributes in v1.8.1.1. The regression is fixed in v1.8.1.6 and
>above.
>
>If upgrading is too hard, I think you can work around it by putting
>"win_build/" instead of "win_build" in your .gitattributes (after
>v1.8.1.6, both should work for your case).
>
>-Peff
> 
>
>