Hi Mox,
Well, I'll add informations. In fact, we need both :
- highlight
and
- cursor implementation
Because the functions used are very similar. For highlighting, I think
Sebastien, Ismael and me have finally found a solution.
What about the cursor ?
After I asked on [EMAIL PROTECTED], Philipp Lohmann and Jan Holesovsky kindly
explained me the missing cursor we were searching for was implemented in
vcl/source/window/cursor.cxx, exactly in ImplCursorInvert() function. In
this function we can see calls to Invert. So once Invert works fine, I
think it should work too for the cursor ( just guessing, and I'll try to
confirm).
And readin more deeply the code inside, it does appear Ismael is right
: we need to take care of all possible cases , and respect as close as
possible vcl implementation. I'm ok with Ismael visoin of the
implementation.
The cursor is based on Invert and they are 3 possible SAL_INVERT =>
let's implement them like that, and see ...
What I'm finally not sure is what exactly means SAL_INVERT_TRACKFRAME ?
(SAL_INVERT and SAL_INVERT_50 are obvious)
Last, I'll try the invert 50 with 50% of opacity. Not sure it works.
Mox Soini a écrit :
Hi all,
Although I haven't gone through code of all the alternative solutions,
The "CGContextSetBlendMode"
-approach seems most appealing to me, because I think we will get the best
performance from it.
Can you please propose code ? Maybe this track is interesting too ?
In fact, I'm like you : I'd have prefered implement Invert( ) and co,
more "Quartzly", but this is yet not possible : we are inside and depend
on everything vcl. :-/
No idea the time/energy it would take to rewrite everything, but maybe
restart from scratch would be a best idea, IMHO ;)
The good thing is we now see more clear with vcl ;-)
I think within vcl/aqua/source/ we should try to use as much native stuff
(Quartz & CG*) as possible and avoid mixing with generic OOo stuff....
I completely agree, and this is what we are doing.
Of course, we could use "::invertRect" as a fallback, if something does not
work, but is there any realistic likelihood for CGContext* not working and
still being able to successfully use "::invertRect" ?
I guess it will be (not sure at all) on Leopard, but just for binary
compatibility, and will lead to strange bugs difficult to analyse. But I
can be wrong.
Eric Bachard
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
