I'd be in favor of a transition to git if Sam and Ryan are up for it,
but not a transition to GitHub. I do use GitHub myself and I do like
it, but SDL is big enough that it kinda needs to maintain autonomy.
Besides, Ryan pointed out some time ago that GitHub has some quirks
he really doesn't like that aren't like other DVCS hosts. There are
git hosting solutions that provide the essential things people want
from GitHub, including ways to do PRs without GitHub actually.
People who use GitHub can pull in a git submodule from somewhere
other than GitHub, so that's not a good reason to give up the SDL
As I said, I no longer regard git and hg as functionally the same
thing with slightly different commands. As I have tried to use and
work with the two more and more, hg just hasn't stabilized the way
git has and thus it has fallen behind for me as I need to work in a
heterogeneous environment where pieces hosted in one DVCS platform
must interact with pieces hosted in another. I can do that with just
about every VCS suite you might use, save for hg. This builds
negative views of hg in my mind.
It ultimately doesn't matter of SDL uses git or hg because I do not
build anything that would bundle SDL right out of hg with my own
code. If I did, it'd irritate me because I probably would feel
constrained by the inability to pull it in via my VCS toolchain
without jumping through a ridiculous number of hoops trying to get
the EXACT supported version of hg installed that works with the exact
version of the import tool. Nonetheless if I had that problem I
would find a solution to it.
But hg's inability to get their act together and stabilize things
enough that we don't need to hot-patch their code in highly breakable
ways to work with their repositories makes me consider it unsuitable
for new projects. Which is too bad because I think hg's commands
make more sense than git's. Nice interface spoiled by implementation
Post by Daniel Gibson
Github has some advantages, like many developers being familiar with
it and already having an account there.
And maybe patches (as pull requests) would be a little more visible
there than in the SDL bugtracker.
But in the end it doesn't matter.
It seems to me like there often is no reaction to bugreports in
bugzilla or bugreports/bugdiscussions on this mailinglist for a long
time, even if there's patches attached (but maybe I'm biased and this
only happens to bugs I'm interested in).
So it seems like the time of the core SDL developers needed to review
bugs and patches is the real bottleneck (which is not surprising, as
they basically do this in their freetime, AFAIK), so attracting more
contributors would not necessarily improve things.
Post by Dominus Post by Dominus
This "use github or I won't/can't contribute" thing amazes me everytime
it occurs in any project.
I think for a lot of people, there is a legitimate concern there. It is
a switch in tools, which can mean time away from working on a game.
Maybe a little time, maybe a lot of time. Maybe it's an unknown amount
of time, in which case, people might be pessimistic, and choose to avoid
the switch altogether.
To note, I'm not suggesting a move of SDL to Github either, but I do
wonder if there are ways to help those who are more familiar with
Github, apply their time and talents to SDL, while also keeping the
signal-to-noise ratio acceptable for SDL's maintainers.
-- David L.
SDL mailing list