[webkit-reviews] review denied: [Bug 20924] [Gtk] Linux/Gtk: Recent tree revisions fail Acid2 and Acid3 : [Attachment 23893] fix regression Acid without remove all changes available on all other platform

[bugzilla-daemon at webkit.org]

Alp Toker <alp at nuanti.com> has denied Mario Bensi <mbensi at pleyo.com>'s request
for review:
Bug 20924: [Gtk] Linux/Gtk: Recent tree revisions fail Acid2 and Acid3
https://bugs.webkit.org/show_bug.cgi?id=20924

Attachment 23893: fix regression Acid without remove all changes available on
all other platform
https://bugs.webkit.org/attachment.cgi?id=23893&action=edit

------- Additional Comments from Alp Toker <alp at nuanti.com>
Mario,

Sorry about my terse reply earlier -- I may not have explained the issue fully.


The change in r35954 needs to be backed out not just because it regressed the
GLib code path but because it's wrong on all platforms: it introduces the
incorrect assumption that all data URLs are Base64-encoded, whereas we're meant
to support non-encoded data URLs as well.

The logic in parseDataUrl() function is pretty messy/difficult to follow but it
was essentially correct for both GLib and non-GLib before r35954 (except for
the loss when going back to a string in the non-GLib code path which you
noticed).

If you study this function (as it was before r35954) you'll see what I'm
getting at and why your latest change isn't correct either. Personally I'd love
to see this stupid function rewritten as it totally obfuscates a simple task.