You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@systemds.apache.org by Mark Dokter <md...@know-center.at> on 2021/02/25 12:27:59 UTC

Re: Git submodules

On 15/11/2020 01:02, Mark Dokter wrote:
> 
> 
> On 04/11/2020 12:25, Janardhan wrote:
>> Hi dev,
>>
> 
> Hi!
> 

Hello again.

A small follow up on this. Better late than never ;-)

>>> Should we introduce git submodules?
>>
>> The git submodules are normally not favoured by the git community[1][2].
>> A while ago, I tried to bring submodule functionality to `git`&`libgit2`
>> projects.
>>
> 
> Thank you for the suggestion! I was not aware of the subtree
> functionality. I might try that :)
> 

I had a go at git subtree and have decided against it as rebasing
becomes troublesome. When I tried rebasing after adding the external
dependency as a git subtree, git wanted to strip away the prefix
directory of the subtree, resulting in conflicts (you'd have to manually
resolve conflicts like Readme.md because everything was moved from
src/main/cuda/ext/jitify to the root of the repository).
There is of course a non trivial workaround for that, but I don't think
it's feasible to expect everybody to be aware of it.

The result is that we stay with a submodule (which I updated).

>> But, subtree is an option worth looking into. Since `jitify.hpp` can
>> just be
>> committed directly as a file or as a subtree. BSD-3 code can be used
>> Apache
>> (needs to be confirmed).
>>
> 
> I actually did not consider legal issues when I started to use jitify.
> In that regard it might be better to stay with submodules for external
> references, so we don't ship them but the users have to pull them in by
> initializing the submodules themselves.
> 

I guess we just need to make sure we include the license file in binary
releases.

Regards,
Mark