Otherwise, in the time it takes the thread to start and other init tasks to
complete, we tend to get an underrun on some systems, which ALSA logs to
stderr.
So this is moved to an InitThread implementation, which runs from the device
thread, right before it begins its main loop.
Reference PR #12632.
(cherry picked from commit ae17b04c0d)
This prevents the waste of an initial buffer of audio on many backends, and is
hopefully harmless on all of them.
Reference PR #12632.
(cherry picked from commit 41636959d2)
This was intended to make the API public, so SDL_hashtable.h got an extreme
documentation makeover, but for now this remains a private header.
This makes several significant interface changes to SDL_HashTable, and
improves code that makes use of it in various ways.
- The ability to make "stackable" tables is removed. Apparently this still
worked with the current implementation, but I could see a future
implementation struggle mightily to support this. It'll be better for
something external to build on top of the table if it needs it, inserting a
linked list of stacked items as the hash values and managing them separately.
There was only one place in SDL using this, unnecessarily, and that has also
been cleaned up to not need it.
- You no longer specify "buckets" when creating a table, but rather an
estimated number of items the table is meant to hold. The bucket count was
crucial to our classic hashtable implementation, but meant less once we
moved to an Open Addressing implementation anyhow, since the bucket count
isn't static (and they aren't really "buckets" anymore either). Now you
can just report how many items you think the hash will hold and SDL will
allocate a reasonable default for you...or 0 to not guess, and SDL will
start small and grow as necessary, which is often the correct thing to do.
- There's no more SDL_IterateHashTableKey because there's no more "stackable"
hash tables.
- SDL_IterateHashTable() now uses a callback, which matches other parts of SDL,
and also lets us hold the read-lock for the entire iteration and get rid of
the goofy iterator state variable.
- SDL_InsertIntoHashTable() now lets you specify whether to replace existing
keys or fail if the key already exists.
- Callbacks now use SDL conventions (userdata as the first param).
- Other naming convention fixes.
I discovered we use a lot of hash tables in SDL3 internally. :) So the bulk
of this work is fixing up that code to use the new interfaces, and simplifying
things (like checking for an item to remove it if it already exists before
inserting a replacement...just do the insert atomically, it'll do all that
for you!).
_GetWinID() doesn't work with keyboard-related BMessages, because Haiku
assumes you know what window has keyboard focus at the time, so these events
don't have a `window-id` property. So when this call failed, the key event
handler would return early.
This was probably a copy/paste error that snuck in at some point, as SDL2
doesn't have this issue.
The comment in the source wasn't true; PipeWire doesn't _have_ to work in
float format. It presumably does if it has to mix, but if a game is the only
thing making noise on the system--a common scenario--then it might be able to
pass, say, Sint16 data straight through to the hardware without conversion.
Fixes#12129.
You can end up with a NULL scratch buffer, which is otherwise not needed on
this path, then ConvertAudio will end up needing that scratch space to move
to float32 to apply gain.
Fixes#12091.
(I assume.)
Audio streams used to accept audio with a src or dest frequency between
4000Hz and 384000Hz. It was arbitrary (or perhaps a relic of older
resampler revisions), and testing shows unnecessary, so remove it.
Fixes#12098.
It used to check for "software" in the hint, but now it has to parse a
comma-separated list.
If it tries to use the software renderer here, you'll end up in an infinite
recursion.
This doesn't affect latency much, but it makes the system usable if the system
drops you down from the bluetooth a2dp profile (headphones) to the handsfree
(I think...?) profile because the bluetooth audio device is also recording,
which would be extremely common in a VoIP app, but also if you're talking
in a different app while also playing audio.
Fixes#8192.
When going into a fullscreen space, update mouseConfinementRect once the
transition is done. When going exclusive fullscreen, force a window sync so
the transition is complete before we do the update.
Fixes#9088.
The Visual Studio projects have been compiling this for awhile, but apparently
we haven't tested this through CMake before. Toolchains without access to
the latest Windows 11 SDK headers have been preprocessing out the dependency
on this code, so we never noticed.
This is just enough to delay MOUSE_LEAVE events until a drag ending outside
the canvas is released, and make sure the MOUSE_UP event has the canvas's
SDL_WindowID, which it wouldn't when mouse focus is was lost by this point.
Fixes#10033.
Any requested swap interval will be saved and set on the first PumpEvents.
Once PumpEvents has been called at least once, swap intervals are set
immediately.
This assumes that events won't be pumped until after an Emscripten main loop
has been defined, and so prevents a warning on the javascript console:
"emscripten_set_main_loop_timing: Cannot set timing mode for main loop since
a main loop does not exist! Call emscripten_set_main_loop first to set one
up."
Fixes#9969.
The test was doing this:
- The output size is 80x60
- The logical size is 40x30
- The viewport is { 10, 7, 40, 30 }
- Draw to fill this whole viewport.
This would offset the filled rectangle a little, as before, but then the
viewport was the size of the entire logical space, so it wasn't a rectangle
centered in the middle of the output, as was expected.
This used to produce the expected output before the fix in
fa7a529912. But it appears the test was
incorrect, so this tweaks the viewport to produce the expected result.
This reverts commit ef758d05c1.
Turns out the bug in #11076 was that we were dropping texture draws
incorrectly, not that scale shouldn't be applied here. The dropped draw calls
were fixed in bf85320947, and this revert is
making the renderer consistent again.
It didn't take scale into account, and the backends would need to do clipping
anyhow, so let the system figure that out for us at the lower level.
Fixes#11318.
Now (only in the generic backend, where it is implemented), this hint is
always respected. Previously it would only be used if no windows were created,
to help reduce CPU load on things like loopwave.
Since it's always used now, the default has changed from 60 (Hz) to 0 (run as
fast as possible). Things like loopwave should still likely force this way
lower than the previous default (and already do: loopwave explicitly sets it
to 5).
The hint can now also be set to "waitevent" which will cause SDL_AppIterate
to only be called after new events have arrived, for apps that are entirely
driven by input and want to consume (almost) no power or CPU time until then.
Fixes#11093.
Fixes#11387.
Ryan C. Gordon