Hello guys,

Since INdT guys released EFL-Webkit port some days ago some hacks
started to appear to test it. Nicolas (captainigloo) did some tests in
order to evaluate using it in enna, Raoul started a cool test
application called ewww:

    svn co http://dev.calaos.fr/calaos-svn/ewww

I did a checkout of this code and created my own git so I keep a
patchset on top of it easily:

   git clone git://git.profusion.mobi/users/gustavo/ewww.git

I diverged a bit with regard to theme, going a more "black & white"
approach and the layout is towards google chrome:
http://staff.get-e.org/~barbieri/ewww.png  (scrollbars are just
indicators and are hidden automatically after 1 second, that's why
there are none in the screenshot).

while doing this I found some problems that I already reported to kenneth:

 - race conditions while painting, efl-webkit should do a lock around
image data while it renders, otherwise we get old results on the
screen, since we have no control over evas buffers, we should just do
these operations when we hold a lock from main thread (possible
request it using an idler, pipe...);
 - valgrind complains a bit.
 - need absolute scroll, or move, so I can implement "Home" and "End";
 - filter keyboard events, possible it's just return bool in
keyPressEvent? problem is that if I press down in a form I don't want
my scrollview to scroll.
 - object deletion while page is loading must be handled properly. if
you use ewww to open a page and close the window, it segfaults.
 - copy & paste, at least text. (low priority)
 - very minor: ewk_webview_object_load_url() should take a const char *
 - export history navigation somehow, since it exports _back(),
_forward()... I want to have access to list and then disable/enable
 - seems that gmail problem is related to frames, it does not work
with javadoc/epydoc, which are really simple html, but uses frames.
 - some way to query for metadata, to present rss link, alternate
css, if it is using cryptography.

My next steps on the interface is to provide proper smart objects for
the view+scroll and then the tab chrome (the current ui can be thought
as a tab). With that I plan to write some basic tabs and then explore
with Google Chrome's idea of multiple process, since I'll not do any
fancy shared memory work, I just plan to reparent
ecore_evas_software_x11 on top of my other canvas, that might do for

If you check the code, you see we still lack lots of resources of a
real browser, but hey! it was a one day work!

Gustavo Sverzut Barbieri
http://profusion.mobi embedded systems
MSN: barbi...@gmail.com
Skype: gsbarbieri
Mobile: +55 (19) 9225-2202

enlightenment-devel mailing list

Reply via email to