[Poppler-bugs] [Bug 64892] very slow render

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=64892

--- Comment #34 from Thomas Freitag <Thomas.Freitag at alfa.de> ---
(In reply to comment #33)
> Thomas the speed improvements are amazing, kudos for the work!
> 
> Bad news is that amongst lots of "minor" changes i've found two changes that
> seem to be regressions in clipping?¿
> 
> On page 7 of bug-poppler13639.pdf, the top box of the second slide used to
> be totally white inside and now has some green on it

Can You compare it rendering it with 150 dpi? You'll see that the green shadow
is also there with the former version. I think that with 72 dpi the pattern
just wastes away. And comparing it with the other boxes: The shadows should be
always there or in neither box. BTW, Acrobat on Windows doesn't show the
shadows in neither box, but I'm not sure if we are just better than Acrobat or
if we have some problems with transparencies. pdftocairo also doesn't show the
shadows, so if You think they shouldn't be there, just open another bug, and
I'll try to find the reason!

> 
> On page 464 of bug-poppler16094.pdf there is now an arrow going down from
> "Interrupt registers & logic" when previously it did start at the edge of
> the bigger box.

Same as ahead: Compare the rendering with 150 dpi, and You'll see the complete
arrow also in the former version. I think that with 72 dpi the pattern just
covers the thin line in the former version, and it begins to appear with 150
dpi. Acrobat on Windows doesn't show the line in the bigger box, even if I zoom
it very high, but this time I think we ARE better: The line makes sense,
instead of pointing just from the bigger box to "Interrupt Mapper", the arrow
from "Interrupt registers & logic" to "Interrupt Mapper" makes much more sense.
But I must admit that once again pdftocairo doesn't show it either. Perhaps
patterns should not have any transparencies? I'm not sure...

> 
> Do you think it's possible to solve these issues? Maybe we can filter
> somehow these files and let them go through the old codepath?

At least with bug-poppler16094.pdf I see no way to distinguish it: The scaling
of the (pattern) images should not be done anyway. A

fprintf(stderr, "%d x %d => %d x %d\n", srcWidth, srcHeight, scaledWidth,
scaledHeight);

in Splash::scaleImage gives:

698 x 268 => 698 x 268
450 x 320 => 450 x 320
498 x 320 => 498 x 320
22 x 24 => 22 x 24
22 x 24 => 22 x 24

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/poppler-bugs/attachments/20130606/4f94f0d5/attachment.html>