Re: unable to push

[david@lang.hm]

On Sun, 1 Jun 2008, Jeff King wrote:

> On Sat, May 31, 2008 at 09:10:26AM -0700, david@lang.hm wrote:
>
>> I created localmaster in an attempt to get things working again. I'll try
>> creating a new master branch... that does make it work again, thanks.
>>
>> what I'm not understanding completely is how I broke things in the first
>> place.
>
> I'm not understanding that, either. You not only created localmaster,
> but you deleted master. Did you maybe do a "git branch -m"?

in the public repo I had never created a 'master' branch. I had just done 
a clone, setup git daemon, and pushed to it.

at some point several days later I found that my push didn't work and that 
I wasn't on any branch in the public repo. in reading the man page for git 
branch I saw the --track option, so I created the localmaster branch to 
try and get things working.

>> going back through .git/logs/HEAD it looks like I did a checkout HEAD^^^
>> and never properly recovered. should I have done a checkout master to get
>> things back on track? I had thought that I could do a checkout of an old
>> version, modify it, do a commit, and the commit would go on top of the old
>> HEAD and I could just keep going (the reason for doing this instead of a
>> reset is that I had already published the intermediate commits to the
>> public repo, so I didn't want to roll back the history, just get the old
>> version of one of the files)
>
> I think you did eventually "recover" which is to say you ended up on the
> localmaster branch. When you did a checkout of HEAD^^^, your HEAD became
> "detached", meaning that you were no longer on any branch, but at a
> specific commit (and there should have been a warning message mentioning
> that).
>
> When you make commits on a detached HEAD, your HEAD remains detached,
> meaning that those commits are not on any branch. So when you did a
> checkout back to some actual branch, those commits were "lost" (meaning
> that no branch contains them, but they are still available by peeking
> through the reflog).
>
> But that shouldn't have had anything to do with the state of the master
> branch; once you detached, you were purely operating on the detached
> HEAD.

that explains things.

David Lang