Cloning and hard links under Windows

[Matt Harbison]

On Mon, 02 Apr 2018 17:10:57 -0400, Bob Hood <bhood2 at comcast.net> wrote:

> Hey, guys.
>
> I'm starting to think ahead to an upcoming role as "Release Manager"  
> where I will be following a modified GitFlow paradigm, and merging and  
> pulling changesets from an integration branch into a release branch.   
> When I attempt a merge from an unknown branch, I will usually personally  
> create a clone of the repository on a separate file system (so it's a  
> complete clone) and then, if the merge fails for any reason, I simply  
> destroy the cloned folder without polluting my own local working copy.   
> If it succeeds, I push upstream, and then delete the cloned folder  
> anyway.
>
> This action of cloning under Windows can take 5-10 minutes on my  
> machine.  For a single instance, that's bearable.  However, if I'm going  
> to be doing this multiple times a day, that time will stack up.

It sounds like you want the ability to have multiple independent  
checkouts, so it seems like the share extension would be useful.

> I tried a clone on the same file system, and it took < 1min to complete,  
> but it is also "linking" instead of copying.  I ran through the cloned  
> working copy, looking for the hard links that Mercurial created, and it  
> looks like some key management files are hard linked to the original  
> repository.  I'm worried that a failed merge would pollute these files,  
> which would mean that my original "pristine" copy would also become  
> questionable even if I subsequently destroy the working clone.

I'm a bit confused.  The only hardlinks should be under .hg/, not the  
working copy.  Nothing under .hg/ should be affected by a failed merge.   
That doesn't happen until the commit.

> My question is: Can I retain the speed of the clone that I get with hard  
> linking (i.e., same file system), and then just make independent copies  
> of the files that are hard linked in the clone to have a completely  
> separate working copy in a timely fashion? I'd really rather not have to  
> use the "--pull" protocol each time I need to attempt merges.
>
> Thanks!
> _______________________________________________
> Mercurial mailing list
> Mercurial at mercurial-scm.org
> https://www.mercurial-scm.org/mailman/listinfo/mercurial