git svn dcommit with a dirty index

Benoit Sigoure· Nov 7, 2007, 19:06 UTC · lore

Hello list,
 From what I understand, when using dcommit, git-svn uses rebase to  
"sync" the history with what has just been committed.  If the index  
is in a dirty state, this will cause trouble.  I thought about using  
git-stash and then git stash apply --index but I'm afraid this could  
be confusing if dcommit actually brings more revision in that the  
ones it has just committed.  I'm not sure this is possible and even  
if it is, it might not be troublesome since if the commits are  
accepted in the SVN repo, they surely don't overlap with commits that  
have been sent in the mean time.  But it's risky, so I don't know  
what to do.  If we use the stash approach, we might want to tell the  
user that we bailed out because of a problem that needs to be fixed  
and that he can recover his changes with git stash apply --index.

Or we should simply check that the index isn't dirty beforehand and refuse to dcommit if it is.

Any suggestion?

PS OT: Eric, have you made any progress on the svn:externals<- 
 >submodules mapping?  I badly need this feature, but I don't want to  
start to work on it if you're currently working on it (or about to  
deal with it) to avoid unecessary effort duplication.

-- 
Benoit Sigoure aka Tsuna
EPITA Research and Development Laboratory

Eric Wong· Nov 11, 2007, 01:42 UTC · re: Benoit Sigoure · lore

Re: git svn dcommit with a dirty index

Benoit Sigoure <tsuna@lrde.epita.fr> wrote:

Show 18 quoted lines

> Hello list,
> From what I understand, when using dcommit, git-svn uses rebase to  
> "sync" the history with what has just been committed.  If the index  
> is in a dirty state, this will cause trouble.  I thought about using  
> git-stash and then git stash apply --index but I'm afraid this could  
> be confusing if dcommit actually brings more revision in that the  
> ones it has just committed.  I'm not sure this is possible and even  
> if it is, it might not be troublesome since if the commits are  
> accepted in the SVN repo, they surely don't overlap with commits that  
> have been sent in the mean time.  But it's risky, so I don't know  
> what to do.  If we use the stash approach, we might want to tell the  
> user that we bailed out because of a problem that needs to be fixed  
> and that he can recover his changes with git stash apply --index.
> 
> Or we should simply check that the index isn't dirty beforehand and  
> refuse to dcommit if it is.
> 
> Any suggestion?

The latter option is much simpler. I actually thought there was already a check in dcommit that prevents it from committing with a dirty index, but apparently not...

> PS OT: Eric, have you made any progress on the svn:externals<- 
> >submodules mapping?  I badly need this feature, but I don't want to  
> start to work on it if you're currently working on it (or about to  
> deal with it) to avoid unecessary effort duplication.

Oops, sorry. I've been busy and forgetful. I'll try to work on it later tonight or tomorrow.

-- 
Eric Wong