object compiled on my laptop that I checked in blows up on my desktop. Yay
for worrying about where you compile. So instead, because we only needed to
call init() to setup jvmpi for garbage collection, only call that if we
touch that functionality so we can keep the same old compiled code kicking
around, nobody that consumes this needs to change things, etc.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@401 ed5b42cb-e716-0410-a449-f6a68f950b19
things to play with garbage collection. The gc stuff required JVMPI which the
"server" jvm doesn't have, so poor saps like me with two processors and gobs
of ram needed to hack things to run yohoho & force java to run with the
"client" vm since it used stuff out of unsafe (but not the gc stuff)
Still need to hit up people to build me the shared objects on other platforms,
but the linux ones are there & working.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@398 ed5b42cb-e716-0410-a449-f6a68f950b19
not actually write out new image to a file if we don't need to.
This cuts my tileset building from 42 seconds to 22 when no files have changed
(as opposed to 4 when it was skipping it altogether, but winding up with a hosed
tsbundles.dat), so it's at least a decent bit better.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@391 ed5b42cb-e716-0410-a449-f6a68f950b19
done some extra logic & changed the tileset on the bundle. I obviously want to fix this
so things aren't so insanely grindy when nothing's changed, but it's time to run for my
train, so in the meantime I'm going to make this burn cpu but do the right thing.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@389 ed5b42cb-e716-0410-a449-f6a68f950b19
Instead, let's directly check the java version number. Oh, the hackery... (The check-versions functionality should perhaps be moved to RunAnywhere in samskivert, but this should work for now.)
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@377 ed5b42cb-e716-0410-a449-f6a68f950b19
enough to cover the scene, I really just wanted a quick & dirty way to get
all objects in the scene.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@372 ed5b42cb-e716-0410-a449-f6a68f950b19
on our sound to play, make sure we do at least our half second fudge, rather
than spewing errors as we try to sleep for negative amounts of time.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@332 ed5b42cb-e716-0410-a449-f6a68f950b19
right now, we only cancel looping sounds after working with a full buffer's
worth of audio, so going too large is problematic.
It'd probably be a Good Thing to do something a little fancier for cancelling
looping sounds more immediately, but for now, we'll stick with close to what's
been the status quo for years.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@331 ed5b42cb-e716-0410-a449-f6a68f950b19
other scenarios to really really be sure. Oh the twisty maze of AWT passages.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@330 ed5b42cb-e716-0410-a449-f6a68f950b19
(e.g. decoding oggs, but not playing wavs)
And then make looping work even if we're playing something that doesn't
support resetting to marks in the stream (e.g. decoding oggs)
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@325 ed5b42cb-e716-0410-a449-f6a68f950b19
how long we think the sound will really take to play, not how long the last
chunk of our buffer will take. With bumped up LINEBUF_SIZE, we'd sleep longer,
but would actually START sleeping much sooner, so we would truncate sounds.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@321 ed5b42cb-e716-0410-a449-f6a68f950b19
given a resource path so that it can be smart about loading from files versus
streams depending on what works based based on where the resources are coming
from. Also moved FastImageIO into com.threerings.resource to avoid a dependence
on com.threerings.media in resource.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@310 ed5b42cb-e716-0410-a449-f6a68f950b19
play oggs), we need to jump through slightly more hoops with our audio.
Also, bumping up LINEBUF_SIZE appears to clear up some general issues with
linux playback being worse than other platforms, but definitely solves issues
where playing ogg turned into a bowl of rice krispies on all platforms.
git-svn-id: svn+ssh://src.earth.threerings.net/nenya/trunk@308 ed5b42cb-e716-0410-a449-f6a68f950b19
Charlie Groves