RFC: Initial git save format

Dirk Hohndel dirk at hohndel.org
Fri Mar 7 08:30:44 PST 2014


On Fri, 2014-03-07 at 08:12 -0800, Linus Torvalds wrote:
> 
> On Mar 7, 2014 5:42 AM, "Tomaz Canabrava" <tcanabrava at kde.org> wrote:
> >
> > can anybody recheck this for me? from what Linus wrote, this should
> work for libgit2  0.20, but I had to extend his macro to check the
> libgitversion to be <= 20.
> 
> You are probably right that it should be <= 20.
> 
> I only tested 0.19 and my self-built libgit tree, and the libgit
> versioning is horrible. The API documentation on the web tries to show
> when the API changed, but when you click on the different version
> numbers to see the actual API for that version, it seems to randomly
> give some random version.
> 
> But the color coding on the API web docs does imply that the change
> happened between 0.20 and current HEAD. So I think <= 20 is OK, but
> breaks HEAD because the version hasn't been incremented to 21 yet.
> 
> But let's aim to primarily support actual releases rather than Git
> tree of the day.

Yes, please. Maybe we can make this something where you can say "use
libgit21 API" on the qmake command line. Let me play with this.

> To make things worse, my googling showed at least one libgit developer
> saying these breakages are OK, because it's not 1.0 yet. People like
> that shouldn't write libraries.

/me gets popcorn...

> Oh well. As Dirk mentioned, we used to have similar problems with
> libdivecomputer. The libgit people probably don't have enough external
> users to see there error of their ways.

Jef has been very responsive since those early days and I think
libdivecomputer has been working well in that respect for us for the
last dozen releases.

If only we knew someone who was influential in the git community.
Someone people listened to. Someone who has that subtle skill of
convincing people with his solid technical arguments.

Oh well :-)

/D



More information about the subsurface mailing list