i have recently moved some of my repos to github, because some
people wanted to contribute via github with pull requests, so i used
hgpull to do it. here's the notes:
*) git repo was 8 times bigger than the hg one after export (git gc
--aggressive fixes this)
*) every merge of a pull request results into a new branch and head
which forces hggit to create another new branch+merge.
*) unrelated heads everywhere
*) i have finally moved to git. git is bloated, but at least it's not python
*) after aggressive gc, git repo is 6MB, while hg is 15MB
I tried once to make a smaller git by reusing the git's source code..
but this seems like an impossible task as long
as the source have big dependencies with itself, so you can't easily
strip down the code. (git is almost 20MB installed).
I discussed in irc about rewriting hg in C, but they just lold and make
fun on it saying i was trolling. those
kind of reactions doesn't happen in #git.
so.. what do you think about creating a suckless git client?
On 07/03/12 15:44, Micheal Smith wrote:
I don't have a strong opinion on this either way. The current setup works for
me. However bitbucket both offers mercurial, and the suckless namespace is
currently free. Might be an easier, and more sensible fit regarding a mirror.
On Jul 3, 2012, at 5:54 AM, Kai Hendry wrote:
On 3 July 2012 12:47, Nick <suckless-...@njw.me.uk> wrote:
Is there any point in this other than as a distributed backup?
I worry it may slightly fragment things, with people searching for
dwm and using the github copy, rather than our primary one. Or
sending us github pull requests rather than patches.
Well I'm hoping for more, since I quite like the github BTS and UI.
Depends on how many of suckless authors I can convince to take an
interest. ;) Don't think it would work if I'm some sort of proxy.
I think the github UI has some value at least.
https://github.com/scklss/dwm/graphs/
I'll work on a @daily cronjob that will hopefully keep it synced.
