The function already returns `ssize_t`, so do not use `errno`
to communicate the reason for failure, instead, return the
negative errno.
`pw_getrandom()` was inconsistent in this regard because
sometimes it simply returned a negative errno without
setting `errno`. This change fixes that as well.
On some filesystems, the directory entry type is immediately
available, so use that to check if the entity is a directory,
and only use `stat()` when the entity type cannot be determined
from the directory entry.
Always use get_avail() and then only fetch the hires timestamp when
enabled to enhance the delay reporting. This way we also recover from
errors from snd_pcm_avail() instead of ignoring them.
This should make the recover after mmap_begin obsolete but we'll leave
that just to be safe.
Document the pulse.fix properties.
Add an option to disable handling of the FIX flags when the pulse.fix.
property is set to an invalid value/0.
See #3317
Buffer sizes smaller than one cycle are possible, so don't assert that.
Instead, just provide as much samples as fits to the buffer.
If we are driver when this happens, emit a warning (once). Similarly to
ALSA, as driver we produce only one buffer at cycle start, and no new
buffers in process. If the whole cycle doesn't fit into the buffer,
recording probably will be broken and we want some debug when there will
be a bug report about that.
`pw_proxy::core` must be initialized for `pw_proxy_init()`
to succeed, so take it as a parameter instead of relying
on the caller to initialize that field beforehand.
`pw_core::core` was initialized to point to itself, it
wasn't changed, and nothing really used it. And the only
user already had a pointer to the core object. So remove it.
dlopen lv2 and sofa plugin modules instead of hardcoding them into the+
filter-chain. This also makes it possible to add more plugin module
types externally.
If we can't find data ports to link, try to find notify/control ports
when making a link.
When applying the link, place the output of the notify as the control
data.
Don't instantiate the graph when negotiated but when starting. This
allows us to get to the target graph rate in case the format rate is
supposed to follow the graph rate.
See #2969
Add ASYNC flag that sets SPA_NODE_FLAG_ASYNC.
This ensure we allocate at least 2 buffers. We need 2 buffers at least
because we don't dequeue/queue a buffer in the process function when
async and we run out of buffers if we have only 1 buffer.
Since UCM doesn't perform the path lookups of alsa-mixer all UCM devices
fall back to the card's properties for e.g. icons and hence usually show
the `audio-card` icon for all ports giving a confusing UI to users as
all ports show the same icon.
As ucm names are pretty standard augment some common port names with
icons.
The old way fails, if a distro has the header but not the library,
which can happen on at least Gentoo with multilib deployments,
where the shared header is present but non-native libraries might not.
This could still fail, if a distro had some but not all libraries for
some architectures but hopefully no one did that. In that case, a compile
test would likely be required via cc.check_header() instead but let's try
the faster fix first.
Reported-by: Sam James <sam@gentoo.org>
Thanks-to: Barnabás Pőcze <pobrn@protonmail.com>
Thanks-to: Xavier Claessens <xavier.claessens@collabora.com>
Signed-off-by: Niklāvs Koļesņikovs <89q1r14hd@relay.firefox.com>
As far as I can tell, the code in question was not doing anything there
and was also breaking the new code elsewhere, so it was put on the
chopping block. Don't mind the squirming tentacles.
Signed-off-by: Niklāvs Koļesņikovs <89q1r14hd@relay.firefox.com>
