still getting 'It is a submodule!" in 1.6.2.5

still getting 'It is a submodule!" in 1.6.2.5

[Tim Olsen]

Hello.  I reported a bug last week [1] with regard to merging a
repository that contains submodules.  To summarize the bug, I was
getting the following error message from git-merge:

fatal: cannot read object 83055ffdddde60d41d9811aae77e78be50b329f8
'rubydav': It is a submodule!

when rubydav was never not a submodule.

Clemens replied with a patch [2], but I don't think it got into git's git.

Junio replied with two versions of a patch [3, 4], of which one of them
appears to have made it into git's git as commit
7dae8b21c2fe39a59661c709f0dc17090dafa5a4

1.6.2.5 was then released yesterday which has commit
7dae8b21c2fe39a59661c709f0dc17090dafa5a4.

So I tried out 1.6.2.5 but I got the same error :-(

To help others reproduce the bug, I have made a clone of our git
repository publicly available.  Here are the steps to reproduce the error:

~/git$ git clone git://nolita.limebits.net/limebits-merge-bug
~/git$ cd limebits-merge-bug/
~/git/limebits-merge-bug$ git branch --track 0.15.x origin/0.15.x
Branch 0.15.x set up to track remote branch refs/remotes/origin/0.15.x.
~/git/limebits-merge-bug$ git checkout 0.15.x
Switched to branch '0.15.x'
~/git/limebits-merge-bug$ git merge origin/bits-tiddlywiki-01
fatal: cannot read object ef0b013adc08cb28a35e790088e55df3fb2d7b15
'server': It is a submodule!
~/git/limebits-merge-bug$

Again, 'server' has never been anything but a submodule.

Thanks for the help!

Tim


[1] http://marc.info/?l=git&m=124094048321413&w=2
[2] http://marc.info/?l=git&m=124099458427695&w=2
[3] http://marc.info/?l=git&m=124103124319607&w=2
[4] http://marc.info/?l=git&m=124103686729847&w=2

Re: still getting 'It is a submodule!" in 1.6.2.5

[Junio C Hamano]

Tim Olsen <tim@brooklynpenguin.com> writes:

> Clemens replied with a patch [2], but I don't think it got into git's git.
>
> Junio replied with two versions of a patch [3, 4], of which one of them
> appears to have made it into git's git as commit
> 7dae8b21c2fe39a59661c709f0dc17090dafa5a4
>
> 1.6.2.5 was then released yesterday which has commit
> 7dae8b21c2fe39a59661c709f0dc17090dafa5a4.

But that is not the one you want.  7dae8b2 (diff -c -p: do not die on
submodules, 2009-04-29) is to allow viewing of such a merge correctly; To
make the merge automatically, you'd need 0c44c94 (merge-recursive: do not
die on a conflicting submodule, 2009-04-29), which is on 'master', but not
on 1.6.2.X (and likely will never be).

Could you be running 1.6.3-rcX instead of 1.6.2.X?  In general, the tip of
the 'master' is always as stable as any released version, if not more.

Re: still getting 'It is a submodule!" in 1.6.2.5

[Tim Olsen]

Junio C Hamano wrote:
> But that is not the one you want.  7dae8b2 (diff -c -p: do not die on
> submodules, 2009-04-29) is to allow viewing of such a merge correctly; To
> make the merge automatically, you'd need 0c44c94 (merge-recursive: do not
> die on a conflicting submodule, 2009-04-29), which is on 'master', but not
> on 1.6.2.X (and likely will never be).
> 
> Could you be running 1.6.3-rcX instead of 1.6.2.X?  In general, the tip of
> the 'master' is always as stable as any released version, if not more.

Thanks Junio for your reply.  We have made a copy of the offending
branch by hand and have managed to merge that instead for now.  I can
wait until 1.6.3 is released.

Thanks again for your help!

Tim

Re: still getting 'It is a submodule!" in 1.6.2.5

[Tim Olsen]

Tim Olsen wrote:
> Thanks Junio for your reply.  We have made a copy of the offending
> branch by hand and have managed to merge that instead for now.  I can
> wait until 1.6.3 is released.

I changed my mind and decided to try out 1.6.3.rc4.  I can confirm that
the bug is fixed.  Thanks!

Tim

______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email 
______________________________________________________________________

Re: still getting 'It is a submodule!" in 1.6.2.5

[Matthias Andree]

Am 05.05.2009, 08:43 Uhr, schrieb Junio C Hamano <gitster@pobox.com>:

> Tim Olsen <tim@brooklynpenguin.com> writes:
>
>> Clemens replied with a patch [2], but I don't think it got into git's  
>> git.
>>
>> Junio replied with two versions of a patch [3, 4], of which one of them
>> appears to have made it into git's git as commit
>> 7dae8b21c2fe39a59661c709f0dc17090dafa5a4
>>
>> 1.6.2.5 was then released yesterday which has commit
>> 7dae8b21c2fe39a59661c709f0dc17090dafa5a4.
>
> But that is not the one you want.  7dae8b2 (diff -c -p: do not die on
> submodules, 2009-04-29) is to allow viewing of such a merge correctly; To
> make the merge automatically, you'd need 0c44c94 (merge-recursive: do not
> die on a conflicting submodule, 2009-04-29), which is on 'master', but  
> not
> on 1.6.2.X (and likely will never be).
>
> Could you be running 1.6.3-rcX instead of 1.6.2.X?  In general, the tip  
> of
> the 'master' is always as stable as any released version, if not more.

Is there an easy-to-find and easy-to-grasp table that lists which branches  
are recommended for which target group? If there is, I have constantly  
missed it.

May I suggest that there be a remark on the download page of git-scm.com  
that "master" is really the recommended branch?

-- 
Matthias Andree

Re: still getting 'It is a submodule!" in 1.6.2.5

[Matthias Andree]

Am 06.05.2009, 11:09 Uhr, schrieb Nanako Shiraishi <nanako3@lavabit.com>:

> Quoting "Matthias Andree" <matthias.andree@gmx.de> writes:
>
>> Is there an easy-to-find and easy-to-grasp table that lists which
>> branches  are recommended for which target group? If there is, I have
>> constantly  missed it.
>
> After each major release Junio sends out a message with a title "Note  
> from the maintainer" (or sth like that).

OK.

>> May I suggest that there be a remark on the download page of
>> git-scm.com  that "master" is really the recommended branch?
>
> I'm afraid you are suggesting to a wrong person; Junio doesn't control  
> what's in that domain.

This was in the hopes that the site maintainer was reading this list, and  
given there's a redirection from git.or.cz to the new site, I deemed it  
was somehow semi-official. At least, git-scm.com is concise :-)

-- 
Matthias Andree