|Subject:||Re: 2.13.40 regtests|
|Date:||Sun, 28 Nov 2010 12:11:31 -0000|
"Phil Holmes" <address@hidden> wrote in message news:...
"Graham Percival" <address@hidden> wrote in message news:address@hiddenOn Fri, Nov 26, 2010 at 07:10:25PM -0000, Phil Holmes wrote:Done this - comparing .39 with .40. I did a pixel-by-pixel comparison, allowing a leeway of 1 in pixel brightness (range is 0It identified 21 files with changes.Wow, I was expecting much more! In that case, this is definitely worth adding. Short-term, I think it's enough if you just use it on your local computer, and use it to identify regressions. In the medium (3-month)-term, we'll look at adding a pixel comparison to the official build process, somehow. Cheers, - Grahamhttp://www.holmessoft.co.uk/homepage/lilypond/imagediffs.htm
I've now re-written the program to be multithreaded - max 4 threads, so 4 instances of LilyPond compiling at the same time. My old version also did logfile checking, but I couldn't get this to work with multiple threads - think each was over-writing the output from the other thread - so I've dropped that completely. The multithreaded version ran in about 25 minutes, generating and comparing about 2000 files. I think the next enhancements will be to state and flag the differences where they are minor (less than 10 pixels, say).
The PC is unusable when it's running - all the Ghostscript command line interfaces get in the way.
-- Phil Holmes Bug Squad
|[Prev in Thread]||Current Thread||[Next in Thread]|