Re: [PATCH] git-p4: Corrected typo.

From: Vitor Antunes <vitor.hda@gmail.com>
Date: Jan 31, 2011, 13:39 UTC
Message-ID: <AANLkTinKQRfwaMpGQiDCkh1RerQ_MrakwvNLAqgmbF9A@mail.gmail.com>
In-Reply-To: <AANLkTinrALdy9w7K50r=k-oMV9V5+7+LKvui_DSmJ6nK@mail.gmail.com>

Hi Thomas,

On Mon, Jan 31, 2011 at 12:51 PM, Thomas Berg <merlin66b@gmail.com> wrote:

Show 9 quoted lines

>> Maybe I'm not seeing some obvious limitation, but I can't imagine a
>> branching structure that can't be imported into git. Could please you
>> give me an example?
>
> Here I was thinking of the fact that git-p4 (last time I checked the
> implementation ) uses the list of branch specs in Perforce for
> figuring out the parent of a branch. Our branch specs have changed
> over time (they are used for different integration purposes), so they
> are no longer usable for this purpose.

My personal git-p4 script uses a configuration option to define the list of branches. I also added an option to get the list of branches filtered by user (p4 branches -u), which avoids waiting for the server since I don't have any branches defined.

> I also discovered bugs in git-p4: in some cases, if the first submit
> to a new branch in Perforce is not identical to the branch it derives
> from, the import was not correct.

Another thing that I modified was the following lines:

1559 for (prev, cur) in zip(self.previousDepotPaths, depotPaths): 1560 for i in range(0, min(len(cur), len(prev))): 1561 if cur[i] <> prev[i]: 1562 i = i - 1 1563 break

This tries to find the root directory of all branches, but does that comparing char by char. So, if you have something like:

//depot/branches/branch //depot/branches/branch_test1 //depot/branches/branch_test2

It will assume that your root branch is //depot/branches/branch, which is wrong. I've modified this to split the string by "/" and compare those items, making sure it will detect //depot/branches as the root directory.

Show 10 quoted lines

> One other issue with Perforce, CVS and many other systems is that they
> branch per file. Therefore Perforce can represent partial merges
> between two branches, which git cannot. Because of this, translating
> merges in Perforce to merges in git is not always possible or
> desirable:
> - if you integrate just one file from one branch to another in
> Perforce, and leave the rest unmerged, you probably want to represent
> it as a normal git commit (not a merge)
> - if you merge almost everything, but leave out a file for some
> reason, you may still want to represent it as a merge in git

Yes, merge detection is something that is working. I never tried to look into this because I don't find it too important for my work flow.

Show 5 quoted lines

> The git-p4raw tool has excellent handling of merges, see details in
> this file around line 4300:
> https://github.com/samv/git-p4raw/blob/master/git-p4raw
> It supports several algorithms for automatic merge detection, as well
> as manually changing it after the import is done.

I'll have to look into this later :)

Bye,

-- 
Vitor Antunes