ogg dev - Apr 2008 - [PATCH] browser_plugin - kate support, build fixes, and misc

Hi,

this patch adds:

 - kate support (what else) - this bumps the Javascript API to 1.1
 - a few build fixes (note that the adding of -L/usr/X11/lib -lX11
looks suspect to me, that shouldn't be needed, so you may want to
leave it out if you think so as well - it's odd as X11 using programs
do configure fine on that box)
 - bad pointer dereference fix (mouse callback)
 - avoid the input widget in test.html firing hotkey'd functions

Note that the rendering code is only implemented for the Linux version
at the moment, and I don't have a Mac or Windows machine, so I'm hoping
a coder with either might be interested in having it working for him/her :)
The callback should work on all platforms though (heavily piggybacking
on the CMML code here).

Note also that this uses simple text from liboggplay, so directionality is not
taken into account (eg, Arabic text displays in the wrong order). This is not
really fixable now anyway as Imlib doesn't seem to support this (if I draw
right to left as a test, characters are flipped too, which we don't want).
This will get fixed once I get my rendering lib done, and liboggplay will have
the ability to send Kate streams as YUV overlay data (which can then be
drawn over a Theora video). This is longer term though (and will add more
dependencies to liboggplay, like cairo and pango).

Only one Kate stream may be selected at a time, but they can be changed
on the fly, with language/category exposed to Javascript so they can be
put in a menu or something. See the new test buttons in test.html.

The font selection is not great - it picks the first font it finds in a list of
predefined paths - though the font can be overridden by Javascript.
I wanted to get the default font from the Mozilla preferences, but it seems
that it's not possible to get to this from a plugin (as opposed to an
extension),
correct me if I'm wrong. First time I've written code for a Mozilla
plugin, so
I'll be missing a lot.

If anyone happens to be interested, I'd welcome comments on the Javascript
API. I'm just thinking it might be a good idea to add text color setters. Or
maybe rendering text twice at different sizes to give an outline so
it's readable
for all backgrounds.

Thanks
-------------- next part --------------
A non-text attachment was scrubbed...
Name: browser_plugin-with-kate.diff.gz
Type: application/x-gzip
Size: 10386 bytes
Desc: not available
Url :
http://lists.xiph.org/pipermail/ogg-dev/attachments/20080428/5bb2a174/attachment-0001.bin

2008/4/28 ogg.k.ogg.k at googlemail.com <ogg.k.ogg.k at
googlemail.com>:
> Hi,
Hi,

I've tried to separate out the general bugfixes from the new
implementation work in this patch:
>  this patch adds:
>
>   - kate support (what else) - this bumps the Javascript API to 1.1
I've not yet committed this to trunk. It does sound pretty cool
though. I don't see any problem with adding pango etc. support, it's
pretty much where we wanted to go for CMML handling anyway.

(btw. I'm tracking the Kate-specific stuff in a local git branch)
>   - a few build fixes
committed in changeset:3570
>  (note that the adding of -L/usr/X11/lib -lX11
>  looks suspect to me, that shouldn't be needed, so you may want to
>  leave it out if you think so as well - it's odd as X11 using programs
>  do configure fine on that box)
thanks for noticing, the configure script wasn't looking for the X11
install location at all. I've updated it to use the autoconf macro
AC_PATH_XTRA, which defines X_CFLAGS, X_LIBS etc. in changeset:3571

Let me know if this works for you or not; eg. it may be necessary to
also put X_EXTRA_LIBS etc. in src/Makefile.am.
>   - bad pointer dereference fix (mouse callback)
I couldn't find that particular hunk, could you please provide a patch
which just fixes this?
>   - avoid the input widget in test.html firing hotkey'd functions
is this specific to the Kate changes? I noticed that it adds a focus
tracking variable, but that seems to be used by some of the Kate UI
also so I'm not sure if it's a general bugfix or Kate-specific.

cheers,

Conrad.

> I've tried to separate out the general bugfixes from the new
> implementation work in this patch:
Sorry, I should have done that, I do tend to fix stuff I find while
coding and neglect splitting up afterwards. Bad me.
> I've not yet committed this to trunk. It does sound pretty cool
> though. I don't see any problem with adding pango etc. support,
it's
> pretty much where we wanted to go for CMML handling anyway.
Ah, good. At the moment, it's still imlib2 only though. Not great
but will do for now. Cairo/Pango will be needed if/when I add my
rendering lib, but I'm just starting on it, so imlib2 it will be for the
near future.

If you have any comments/concerns about the Kate code, feel free
to comment. Especially on the Javascript API side (never done any
of this before).
> Let me know if this works for you or not; eg. it may be necessary to
> also put X_EXTRA_LIBS etc. in src/Makefile.am.
will try this.
> >   - bad pointer dereference fix (mouse callback)
>
> I couldn't find that particular hunk, could you please provide a patch
> which just fixes this?
Two hunks, the first one to initialize the mouse callback pointer
to NULL, and the second one to test for NULL before actually using it:

--- svn/annodex/browser_plugin/src/plugin.cpp   2008-04-04
15:14:28.000000000 +0100
+++ svn/annodex/browser_plugin-with-kate/src/plugin.cpp 2008-04-27
22:08:08.000000000 +0100
@@ -245,7 +244,8 @@
   mCmmlCallback(NULL),
   mAsyncCmmlCallback(NULL),
   mEndPlayCallback(NULL),
-  mPlaylistCallback(NULL)
+  mPlaylistCallback(NULL),
+  mMouseClickCallback(NULL)
 #if defined(XP_MACOSX)
   ,
   mOutputCleared(FALSE)
@@ -1037,7 +1175,9 @@

   if (mMouseButtonDown) {
     mMouseButtonDown = FALSE;
-    mMouseClickCallback->Call();
+    if (this->mMouseClickCallback != NULL) {
+      mMouseClickCallback->Call();
+    }
   }

   if (useSemaphore) {

> >   - avoid the input widget in test.html firing hotkey'd functions
>
> is this specific to the Kate changes? I noticed that it adds a focus
> tracking variable, but that seems to be used by some of the Kate UI
> also so I'm not sure if it's a general bugfix or Kate-specific.
Well, not really, more like only useful for the new kate test functions:
the keypress routine would be called for every keypress, but all
existing test routines would expect numbers in the input box, and no
numbers were hotkeys for any function, so nothing would get done.
Since I added a function (to see the font to use) that expects input
with other characters than numbers, some test functions would be
invoked.

However, looking at the patch, the input_focus variable isn't used
by the Kate UI. Were you talking about something else ?

Thanks for your time.

> >   - kate support (what else) - this bumps the Javascript API to 1.1
>
> I've not yet committed this to trunk. It does sound pretty cool
Can you let me know the reasons (eg, wanting to test first, concerns
over the substance of the changes, or other) ?

There were no comments so far about the changes to the Javascript
API, so I'll assume people have no problem with those ?

Cheers