I think the main barrier to pushing this is not so much the pipeline code
itself,
and whether it works 100% (since it would be off by default) but
making sure it is architecturally sound and doesn't break anything that exists.
I have a very few superficial comments from my quick look at a few files
that are *changed*.
X11SurfaceData and the glyph caching changes need a slightly longer
look. I do note we'd need to make some closed source changes to
match some of it.
1) src/share/classes/sun/font/FontManager.java
shows that this isn't synced with Roman's refactoring.
Shouldn't be a big deal though as you only change
one line and I don't think even that line should be changed
where you changed an array size from 20 to 50 :
>private static CompositeFont [] compFonts = new CompositeFont[50];
I can't guess how x render defines new logical fonts.
2) src/share/classes/sun/java2d/pipe/AAShapePipe.java
imports sun.java2d.pisces.*
but doesn't (and shouldn't) do this directly, and
also has new CairoRenderingEngine(); //.getInstance();
I'm not sure what's intended here. Obviously it
shouldn't directly instantiate that here, but also
I don't even see this class in the webrev.
3) src/share/classes/sun/java2d/pipe/RenderQueue.java
the size change from 32000 to 12000 probably should be explained :
75 private static final int BUFFER_SIZE = 12000;
4) src/share/classes/sun/java2d/pisces/PiscesRenderingEngine.java
This is just adding white space ..
5) The makefiles introduce a compile time dependency on libxrender.so
which I believe should be fine for any *nix platform we'd consider
supporting in JDK 7+ and should also be OK on our build platforms.
6) src/solaris/native/sun/awt/awt_GraphicsEnv.c
What was wrong with using the #define here ?
969 if (xerr->minor_code == X_ShmAttach) {
->
969 if (xerr->minor_code == 1) {
-phil.
Clemens Eisserer wrote:
Hello again,
I am sorry, but currently I can't spend a lot of time working on the
integration of the xrender pipeline.
I am well aware the code isn't production ready, and sometimes just
plain ugly ;)
There are some global changes, most touching glyph handling, because
the xrender glyph cache logic is written in plain java, and e.g. it
wasn't possible to get notifications of glyph disposal on pure java
code as far as I have seen.
It would be cool to get comments on that stuff first, the other things
could follow piece by piece later.
I did some changes in the OGL pipeline, but I don't have windows, so
D3D changes are missing, as well as the proprietary scaler
modifications.
The webrev is located at: http://93.83.133.214/glyph_changes_001.zip
The changes are in:
StrikeCache.java
fontscalerdefs.h
freetypeScaler.c
sunFont.c
OGLTextRenderer.c
GlyphDisposedListener.java
The only overhead I can see if the additional locking in freeXMemory,
the structure itself should remain at the same size (using otherwise
unused padding).
Any idea wether this requirement could be solved more elegant / efficient?
Thank you in advance, Clemens
