[PATCH weston 00/10] Implement screenshot-based testing with the headless renderer

[Bryce Harrington]

On Wed, May 13, 2015 at 02:49:47PM +0100, Daniel Stone wrote:
> Hi,
> 
> On 13 May 2015 at 13:19, Bryce Harrington <bryce at osg.samsung.com> wrote:
> > Thanks both of you for the reviews.  I've implemented some of the
> > changes you suggested:
> >
> >  * Refactor cairo out of the test client backend code entirely
> >    by utilizing the weston test surface structure to carry the
> >    specific data that the check_ routines need.
> >  * Drop refactoring redundant pixel copying code to shared
> >  * Take stride and bpp into account in check functions
> >  * In test case, set the bpp for the screenshot
> >  * Fix screenshot protocol description
> >  * Convert C++ style comments to C
> >  * Don't leak screenshot object
> >  * s/ec/compositor/
> >  * Instead of spinning on client_roundtrip, check for client's wl_display
> >  * Drop unneeded wl_surface in test case
> >  * Restore missing weston_log in idle_launch_client()
> 
> Gosh, thanks!
> 
> > Still needing to be done:
> >
> >  * Split up test case code into several functions
> >
> > You suggested a lot of other changes, which I've not done, but did spend
> > some time investigating:
> >
> >  * Split the capture screenshot stuff out to its own interface, or
> >    alternatively, use wl_callback to signal the completion.
> >
> >    I did take a shot at this, but the code just kept getting more and
> >    more broken so I cut losses after a few hours banging my head on it.
> >    Code seems to work fine as implemented.
> 
> Fair enough. Should be pretty easy to fix; if you want to post a WIP
> to do this as a followup, I'm sure you'd find someone to walk through
> and figure out what was wrong. I think having this in 1.8 is more
> important than having it literally perfect first go around.

Maybe I'd have better luck on a second try.  Part of the problem is in
this section of code everything's named either 'test' or 'weston_test'.  %-)
 
Bryce