[Webkit-unassigned] [Bug 26379] Reconsider image decoding architecture/APIs

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=26379





------- Comment #18 from zecke at selfish.org  2009-06-14 15:26 PDT -------
Unforking Chromium meant breaking Gtk+, WX, EFL, Torchmobile Windows CE. The
cost of you "unforking" is that _now_ every port needs to copy and paste (fork)
+115 lines of code. It is not a fair trade.... there is no reason to create
this mess.

Performance: As with every thing it depends. For Skia and even the current
Cairo usage I agree that the additional malloc, memcpy, free create problems in
terms of memory copying and memory fragmentation and still in some cases this
doesn't matter... E.g. you might have a memory strategy where you keep the
compressed image in memory and move the uncompressed image to another
process/system (e.g. GL, or Xserver memory) or you have API that can operate on
the allocated memory without a memcpy.

And I strongly discourage forcing every port to copy and paste the exact +115
lines of RGBA32Buffer.cpp for absolutely no gain about the status quo of
yesterday. It is no acceptable solution to force this duplication on ports.


Your changes have so many conceptually weaknesses that you make every other
port suffer and this needs to change. I see you have a hardtime admitting that
there is _no_ improvement for cairo, wx, EFL, yet. Instead you talk about a
mystique future improvement... Copy and Pasting RGBA32Buffer is not the future.
You should evaluate how to adjust the ImageBuffer interface and make
ImageDecoder truly corss-platform again bringing your improvement to everyone
with client side image memory.

Until then I strongly urge to unbreak the damage that was done to RGBA32Buffer.


-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.