[postgis-devel] repository hosting (was: is PostGIS ready for git ?)

Sandro Santilli strk at keybit.net
Thu Oct 15 05:41:19 PDT 2015


[subject changed to not mix git and GitHub]

On Thu, Oct 15, 2015 at 01:05:03PM +0100, Mark Cave-Ayland wrote:

> For our purposes, the github pull requests are roughly equivalent to
> someone waving a flag and shouting "hey, go look at this patch!" rather
> than as a direct pull request. Perhaps we should think about integrating
> the buildbots on a "pending" branch that committers can then merge into,
> and then once a clean set of builds have passed can be merged and pushed
> to master. Otherwise it doesn't really save the developer much work as
> he/she would have to manually add the remote, download the code and run
> the regression tests before a commit and then push. Is this something
> that we can do with github as-is?

Partial automatic checking of pull request is already in place:
https://travis-ci.org/postgis/postgis/pull_requests
A buildbot (Travis) runs tests for some configuration.
But only for code that gets in GitHub branches.

Regina's buildbots (Jenkins) could be configured to also test
specific branches on GitHub or anywhere else.

More buildbots are available (but unconfigured) on the GitLab
mirror, see https://trac.osgeo.org/postgis/ticket/3209.

Manually merging pull requests isn't hard, you can just append .patch
to the PR url and you obtain a patch including author and commit log,
and it can be applied with "git am". I actually often use a script
I wrote for accepting patches sent via URL on the Gnash mailing lists:
http://strk.keybit.net/tmp/slurp (Usage: slurp <url>).

> I'm mildly hesitant about putting all our eggs in the github basket for
> 2 reasons: firstly everyone thought sourceforge was a good place to host
> their code until they changed their policy re: adverts and orphaned
> project binaries - hopefully this will never happen, but it's something
> to think about when comparing with osgeo hosting. Secondly I've been
> using github quite a lot for GSoC work over the summer and I don't know
> whether it's just being based in the UK or the strange hours involved,
> but during that time I've experienced several outages of more than
> several minutes which is frustrating when you're trying to help a
> student upstream patches.

Thanks for your comments.

--strk;



More information about the postgis-devel mailing list