[GSOC] Tiled Rendering, ahunt, Weekly Report 07+08 / WK27+28

Andrzej Hunt andrzej at ahunt.org
Mon Jul 14 04:46:10 PDT 2014

Hi All,

Unfortunately I forgot to send my status last week, hence this is the 
combined summary for the last 2 weeks:

=== 07 / WK27 ==
* Worked on moving Calc's column/row-bars to using Virtual-Device
   scaling rather than pixel coordinates, meaning they now match.
** Still some issues with scrolling, probably the various invalidate
    and update calls mixing pixels with document coordinates.
* Implemented a quad-tile test widget.
** Used for the gtk test app.
** Planning to move this into a separate widget, and allow selection
    within the gtktiledviewer to test the production widget or the
    quad-tile widget.
* Fixed border transition issues -- just more confusion with
   Rectangle's definition of its boundaries. (Affects the svp backend.)
   [ See writer_quad.png for ugliness around the border transition, now
     fixed. ]

=== 08 / WK28 ==
* Impress Tiled Rendering: implemented.
** Hard to find where we can plug tiled rendering in, but once found
    wasn't much work to implement.
** Foreground images not rendered yet (same thing happens in Calc -- is
    a shared bug AFAICT).
    [ See impress_noimage.png ]
** No switching between slide/notes/both yet (currently defaults to
    last used mode).
* Added part selection to gtk tiled viewer (and cleaned up the API
   for both LOK and the tiled widget).
   [ Dropdown menu, also visible in impress_noimage.png ]
* Basic Tiled Rendering test implemented.
** Only partially enabled -- still some issues with viewports in the
    drawing layer -- content of document matches except for the page
    edges -- however a tentative viewport fix also fixes this (but I
    don't yet understand the code, so won't push that to master until
    I know what's actually going on -- once done the unit test can
    be fully enabled / extended to sd (and later sc)).
** [ See test_assembled.png to see a larger tile assembled from 4
    sub-tiles -- when drawing the same area as one larger tile, the page
    edges all look like they do in the top-left sub-tile -- with my
    temporary/branch-only viewport patch all the edges are now drawn
    consistently, as visible in the sub-tile pasted on top. ]
* Merged Impress work + tests + viewer improvement to master (and also
   rebased the calc work on top of that, now in

* Modified gtktiledviewer to allow selecting between the normal
   and quad-tiled widget.
** (We could use the gtk/glib inheritance mechanisms to make the
    quad-tile widget extend the normal widget (and thus retain
    full functionality), however it seems wasteful to add
    virtual methods etc. to the widget just for this purpose, instead
    the quad-widget is a duplicate, albeit with some features now missing
    (i.e. no real point in duplicating new features from the usual
    widget). Moreover, the main widget is (hopefully) likely to evolve
    away from rendering of a large fixed tile -- whereas keeping the quad
    tile widget may be of some use even if the main widget implementation
-------------- next part --------------
A non-text attachment was scrubbed...
Name: writer_quad.png
Type: image/png
Size: 92303 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/libreoffice/attachments/20140714/4ced3d40/attachment-0003.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: impress_noimage.png
Type: image/png
Size: 50407 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/libreoffice/attachments/20140714/4ced3d40/attachment-0004.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test_assembled.png
Type: image/png
Size: 65926 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/libreoffice/attachments/20140714/4ced3d40/attachment-0005.png>

More information about the LibreOffice mailing list