Re: [jgit-dev] RefDirectory/LockFile: preserve .git/packed-refs symlink

[Marc Strapetz <marc.strapetz@xxxxxxxxxxx>]

On 13.02.2018 00:31, Matthias Sohn wrote:
> On Mon, Feb 12, 2018 at 1:34 PM, Matthias Sohn <matthias.sohn@xxxxxxxxx 
> <mailto:matthias.sohn@xxxxxxxxx>> wrote:
>
>     On Mon, Feb 12, 2018 at 11:31 AM, Marc Strapetz
>     <marc.strapetz@xxxxxxxxxxx <mailto:marc.strapetz@xxxxxxxxxxx>> wrote:
>
>         I've a pending patch which resolves a bug and which I think is
>         ready to be merged:
>
>         https://git.eclipse.org/r/#/c/114800/
>         <https://git.eclipse.org/r/#/c/114800/>
>
>         and a related patch regarding symlinks:
>
>         https://git.eclipse.org/r/#/c/114799/
>         <https://git.eclipse.org/r/#/c/114799/>
>
>         for which I have addressed the problem in a more general way in
>         (still needs feedback):
>
>         https://git.eclipse.org/r/#/c/114810/
>         <https://git.eclipse.org/r/#/c/114810/>
>
>
> can you please comment on Jonathan's comments in
> https://git.eclipse.org/r/#/c/114799/2/org.eclipse.jgit/src/org/eclipse/jgit/util/FileUtils.java 
> ?

Sorry, I forgot about the dependency between these patches. I've added 
comments. After Jonathan had suggested to change the LockFile itself, I 
had focused on this crucial change first.

> Are changes 114799 and 114819 alternatives or do you think we need both ?

The LockFile patch (114810) would be an alternative to 114799 and 
partially to 114800.

> I'd prefer if the commit messages are self-contained, they should 
> capture the important
> facts. Referring to a bug having additional information is ok but the 
> commit message should
> explain essentials necessary to understand the change and the reasons 
> why it's needed.

I'm going to add details from the bug description once it's clear which 
route to follow. In case of the LockFile-route, would you consider to 
include this in 4.11 or is it something for 5?

-Marc