Re: How do you hack on GNOME? How can we do better?

[Owen Taylor <otaylor redhat com>]

On Tue, 2015-07-21 at 08:38 -0400, Colin Walters wrote:
> On Mon, Jul 20, 2015, at 10:40 PM, Owen Taylor wrote:
>
> > And some of the things that were done to make gnome
> > -continuous robust - like assembling a build root from scratch for 
> > each
> > build and building each module from scratch - make it much slower 
> > and
> > more resource intensive for local hacking than jhbuild.
>
> I think buildroots quite fast to construct because they use 
> hardlinking rather than the IO intensive traditional package manager 
> install.  On my home server (SSD) it's 5-10 seconds - and most 
> importantly, it's cached - if none of the previous components 
> changed, it's directly reused.

There's no doubt that the buildroot construction in gnome-continuous is
*fast for buildroot construction*, but for local hacking, 5-10 seconds
plus the time to build the component from scratch is a pretty long
cycle time.

Right now, I feel that the bar for when gnome-continuous in a VM is the
easiest way to do things is quite high. I might do it if I had to work
on GDM, but I can't imagine recommending it to a newcomer to GNOME who
wanted to fix a few bugs in gedit.

Do you have a sense in your head about whether it would be possible for
someone to enhance gnome-continuous so that really is suitable for that
"fix a few bugs in gedit" case? If we can leverage the builds that are
being done on build.gnome.org, we potentially could get away from the
the 12 hour yocto builds, the 2-hour webkit builds, and the 30-40GB of
disk space. 

- Owen