This sets up close pipe and socket internally, and exposes the resulting pathname socket and close_fd cleanup as an io.Closer.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This is not guaranteed to have completed after a roundtrip. This is leftover from when Roundtrip also sent and waited for sync.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This should not be handled on every receive as it could cause valid (though impossible in current upstream implementation) messages to be rejected and raise a protocol error.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This change fixes handling of non-fatal errors during a roundtrip as there can be multiple receive calls per roundtrip.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This name is more correct since it does not roundtrip, but receives messages. This is also more consistent with the method on event proxies.
Signed-off-by: Ophestra <cat@gensokyo.uk>
The PipeWire protocol does not work with Go abstractions. This change makes relevant methods call sendmsg/recvmsg directly.
Signed-off-by: Ophestra <cat@gensokyo.uk>
These are PipeWire spa_dict keys. Interestingly many keys in the sample are undefined and appear as magic strings in upstream source code.
Signed-off-by: Ophestra <cat@gensokyo.uk>
These are mostly small formatting changes, with the biggest change being to UnexpectedEOFError where its kind is now described as part of the error type.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This should make things easier to navigate, and possible to fully automatically generate the constants in the future.
Signed-off-by: Ophestra <cat@gensokyo.uk>
These errors are recoverable and should not terminate event handling. Only terminate event handling for protocol errors or inconsistent state that makes further event handling impossible.
Signed-off-by: Ophestra <cat@gensokyo.uk>
These proxies (with special cases documented in the implementation) are only safe for use after acknowledgement from the server.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This consumes the entire sample, is validated to send identical messages and correctly handle received messages.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This was written when the protocol was still barely understood, so none of the types here are correct and match the rest of the protocol. This change corrects these types.
Signed-off-by: Ophestra <cat@gensokyo.uk>
I could not get the server to produce these events, however I am confident enough with the implementation to do it by hand.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This removes the need for manual splitting. The understanding of the format is robust enough to allow this to happen anyway.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This is still not efficient by any means, but it should eliminate most non-reflect allocation (all allocation if PODMarshaler is not used).
Signed-off-by: Ophestra <cat@gensokyo.uk>
The message in the sample does not correspond to any known method call. The spec does not mention what to do with messages like this, but all existing usage code simply drops it.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This appears to add *one single entry* compared to the message before it. The inefficiency of this protocol is beyond imagination.
Signed-off-by: Ophestra <cat@gensokyo.uk>
The POD itself is serialised without requiring a special case, however its presence is only indicated by the difference in size recorded in the header and payload.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This performs the check within the bounds of the POD only. This was not caught since spa_dict was only used as the final struct field until now.
Signed-off-by: Ophestra <cat@gensokyo.uk>
Once again, already entirely supported, the offset is not yet fully verified but makes intuitive sense. Will verify this on future occurrences of the message.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This is a terrible type that defies the type system. It is implemented on the concrete type to avoid special cases.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This is still NUL terminated strings, and an extra NUL character on an 8-byte string does cause an extra 7 bytes of padding.
Signed-off-by: Ophestra <cat@gensokyo.uk>
Test cases are from interactions between pw-container and PipeWire. Results are validated against corresponding body.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This condition causes two runs to be created on a pull, as gitea does not check whether a run has already been created for the current commit.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This significantly increases readability. This patch is pretty big so it is being done after mostly everything has settled.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This interface is getting replaced, so relocating it to the deprecated wrapper package before working on its replacement.
Signed-off-by: Ophestra <cat@gensokyo.uk>
Note: wl_registry_add_listener is undocumented everywhere. Its implementation calls wl_proxy_add_listener which returns 0 on success or -1 on failure.
Signed-off-by: Ophestra <cat@gensokyo.uk>
The sandbox which ldd(1) runs in does not inherit parent work directory, so relative pathnames will not work correctly. While it is trivial to support such a use case, the use of relative pathnames is highly error-prone and generally frowned against in this project. The Exec function remains available under the same signature until v0.4.0 where it will be removed.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This prevents blocking from failures caused by ldd(1) emitting output that is not anticipated by the decoder.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This is mounted for the default read-only /dev/ when programs want to use shm_open(3). Defining it here is less error-prone and saves the extra append at runtime.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This is cleaner and makes more sense. The longer LDFLAGS was never a valid concern since it is always inserted by a script.
Signed-off-by: Ophestra <cat@gensokyo.uk>
These packages are highly specific to hakurei and are difficult to use safely from other pieces of code.
Their exported symbols are made available until v0.4.0 where they will be removed for #24.
Signed-off-by: Ophestra <cat@gensokyo.uk>
This package is ugly and is pending removal only kept alive by xdg-dbus-proxy.
Its exported symbols are made available until v0.4.0 where it will be removed for #24.
Signed-off-by: Ophestra <cat@gensokyo.uk>
While fd created from this side always has the FD_CLOEXEC flag, the same is not true for files left open by the parent. This change prevents those files from leaking into the container.
Signed-off-by: Ophestra <cat@gensokyo.uk>
cat