[patch] inlines

[Germain Garand]

Le Mardi 10 Février 2004 19:53, Stephan Kulow a écrit :
> Am Dienstag 10 Februar 2004 20:36 schrieb Leo Savernik:
> > I'd say commit this to HEAD soon, and if Dirk and Coolo agree, backport
> > it.
>
> I have 614 tests with changed renderer tree output, so figuring alone if
> everything is harmless offset shift will take a while. 

Sure ;(
But actually, if you take off the RenderInline changes which aren't 
significant, it's rather ~ 300 testcases. 
I have also toyed a bit with the possibility to compare positions and only 
account for changes > e.g 2pix. 

See attached script (meant to be to run on a log of "diff -ru 
regressions/baseline regressions/output", invoked as e.g:

$ diff_regressions.pl  -r RenderInline -t 3 log.diff

will skip lines with ^[-+]\s*RenderInline and eventually remove the file from 
output if it's the only diff for it, then take a 3 pixels threshold for 
keeping or throwing diffs in XxY coordinates provided the rest of the line is 
identical.)

Here, I got 316 changes without RenderInlines, and 131 changed testcases with 
a 2 pixel threshold

That's with a week old test_regression... 
I tried updating, but now test_regression aborts, rumbling about unicode fonts 
that I thought I had installed ;(

(when I do a "xfd -fn 
'-adobe-helvetica-medium-r-normal--12-*-75-75-p-*-iso10646-1'", I have
 -adobe-helvetica-medium-r-normal--12-120-75-75-p-67-iso10646-1 displayed, 
isn't that good enough?!)

> I guess I better
> implement the screenshot comparision soon so we have a better way to
> tell what tests actually changed.

I'm not convinved this would be much more significant than a pure textual 
compare but I can't wait to see the proof of concept, just for the hell of it 
:)

Germain
>
> Greetings, Stephan




-------------- next part --------------
A non-text attachment was scrubbed...
Name: diff_regressions.pl
Type: text/x-perl
Size: 2034 bytes
Desc: not available
URL: <https://mail.kde.org/mailman/private/kfm-devel/attachments/20040211/3631a5f0/attachment.pl>