In JDK6, there are 4 abstract methods added to java.nio.Buffer. This
patch adds these. This has no impact on any of the subclasses because
they already implemented all of these 4 methods.
2007-04-12 Roman Kennke <[EMAIL PROTECTED]>
* java/nio/Buffer.java
(array): New abstract me
This fixes CharSequence wrapping CharBuffers by using a specialized
CharBuffer subclass for CharSequences. The previous implementation had
some problems, most importantly it copied the original CharSequence,
where it should provide a read-only CharBuffer on the original
CharSequence, and it was inc
And this is part two of the fix: optimized drawing is disabled in
CairoGraphics2D if any glyphs have been transformed. The fallback code
path respects transformations, while the optimized code doesn't.
Cheers,
Francis
2007-04-12 Francis Kung <[EMAIL PROTECTED]>
PR 23887
*
Hi,
The first of two patches to address a long-standing bug with the text in
vertical progress bars... this one lays out glyphs properly in the
GlyphVector, and fixes spacing/alignment in the progress bar.
Cheers,
Francis
2007-04-12 Francis Kung <[EMAIL PROTECTED]>
PR 23887
Hi Mark,
Very nice. In combination with my FontPeer-leak-workaround this seems to
be pretty solid. In what cases are you now still seeing leaks?
The two open bugs - PR 31539 (ProgressBar) and 28186 (drawString) - were
still leaking pretty badly after my patch. Your FontPeer cache fixes
3153
Hi,
This patch caches the Freetype font map, rather than creating &
destroying a new one for each font peer we create. This enables Pango's
internal caching and mostly fixes the ProgressBar memory leak.
Cheers,
Francis
2007-04-12 Francis Kung <[EMAIL PROTECTED]>
PR 31539
Hi all,
This commit reorders the parts of GregorianCalendar.computeTime() that
decide which fields to use into the priority order outlined in
http://java.sun.com/javase/6/docs/api/java/util/Calendar.html#date_resolution.
The previous incarnation was in an odd order and had some fun fallback
cases.
Hi Francis,
On Wed, 2007-04-11 at 17:10 -0400, Francis Kung wrote:
> While trying to fix some other font-related memory leaks, I came across
> a few issues. They don't plug the major leaks we're seeing, but are
> still worth fixing...
Very nice. In combination with my FontPeer-leak-workaround
On Thu, 2007-04-12 at 16:44 +0200, Mark Wielaard wrote:
> It is correct as far as I can see. So please do commit.
> Do you actually have so little LocalRefs that you notice this in cacao?
This patch is more than a year old, the local references table support
back than was... not existent :-)
>
>
On Thu, 2007-04-12 at 16:22 +0200, Mark Wielaard wrote:
> On Wed, 2007-04-11 at 23:44 +0200, Christian Thalinger wrote:
> > This patch checks for the number of arguments of the gethostbyname_r
> > library call. The six- and five-argument case is implemented, the
> > three-argument case is missing,
Hi Christian,
On Wed, 2007-04-11 at 22:26 +0200, Christian Thalinger wrote:
> I just found this one in my local working copy. Should I commit it?
It is correct as far as I can see. So please do commit.
Do you actually have so little LocalRefs that you notice this in cacao?
Could you please writ
On Wed, 2007-04-11 at 23:44 +0200, Christian Thalinger wrote:
> This patch checks for the number of arguments of the gethostbyname_r
> library call. The six- and five-argument case is implemented, the
> three-argument case is missing, but I added an #error.
>
> Ok to commit?
Yes thanks. I'll als
Hi,
On Wed, 2007-04-11 at 23:33 +0200, Christian Thalinger wrote:
> MIPSpro warned about that. Commited.
>
> 2007-04-11 Christian Thalinger <[EMAIL PROTECTED]>
>
> * native/jni/native-lib/cpproc.c (cpproc_forkAndExec): Fixed
> warning on some compilers, e.g. MIPSpro.
What a lame c
Hi Christian,
On Wed, 2007-04-11 at 23:37 +0200, Christian Thalinger wrote:
> IRIX had a problem with including sys/fcntl.h directly. Commited.
Thanks, also added to the release branch now.
Cheers,
Mark
signature.asc
Description: This is a digitally signed message part
Hi,
this was the simplest fix (actually workaround) for the memory leak we
were seeing with the FontPeers. It just keeps a least recently used
cache of AffineTransform mappings to TransformAttributes to use in the
font maps. That way we use the same fontpeer whenever all attributes are
the same (w
Hi,
On irc Christian pointed out that we use some cairo functions only
introduced since 1.1.8. So this configure check makes sure that version
or something higher is installed on the system.
2007-04-12 Mark Wielaard <[EMAIL PROTECTED]>
* configure.ac (COMPILE_GTK_PEER): Add check for cairo
16 matches
Mail list logo