Pipewire 1.0.7 #78
Open
Pipewire 1.0.7 #78
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
When a memblock closes its fd, have it emit signal to other memblocks possibly sharing the fd, so they know it is now invalid. When a memblock has an invalidated fd, it'll not allow mapping more memory, nor match against fd searches. This avoids bugs where kernel fd reuse causes already invalid memblocks to be used. Generally, higher level code should make sure memblocks get freed in right order so that this situation does not occur, but try to keep mempool in consistent state regardless.
Props set on the node directly should also result in emission of changed notify when required, just like we do for params set on the port. This accidentally used to be done because stream_update_params() unconditionally used to emit this for us. But commit 94cde30 changed that. Fixes #3833
The remote end may destroy the port via client_node_port_update(), before corresponding pw_impl_port_mix are released. clear_port() removes all struct mix, but this prevents the pw_impl_port_mix from being removed from io_map, which causes stale mix ids be left in io_map, so we end up continuously allocating new io areas. Make lifecycle of io_map entries match port_init_mix/release_mix exactly, separately from the lifecycle of the port and struct mix. When freeing struct mix in port_release_mix(), make sure it corresponds to the mix being released.
Avoid leaking buffers when freeing mix, in case the port was not cleared properly. These leaks don't seem to be occurring currently, but better be sure.
When the server closes the connection in non-interactive mode, we need to set the quit flag to avoid going into second mainloop_run() that will just block forever. See #3837
These functions were using the object serial as the object id
Add a MAPPABLE data flag that hints that the fd in the data is mappable with a simple mmap/munmap. Normally, DmaBuf is not mappable like that unless explicitly indicated with this flag. Set the MAPPABLE flag on the DmaBuf from v4l2 and libcamera fd. When asked, mmap the buffer memory in all cases when the MAPPABLE flag is set. This solves the case where v4l2 has exported DmaBuf and is streaming to node A and then node B links but doesn't get automatically mmaped memory. Fixes #3840
When we remove a global, we set NULL,NULL,NULL as its metadata. This then results in the implementor of the metadata to emit the property change event for the subject. When we decide to propagate this event to the listeners we not only need to check if the client can see the global but also if the global was removed and this is a remove metadata event. Fixes metadata remove events when a global is removed.
impl-metadata would not monitor for globals being removed. This would cause stale metadata to remain in the store, causing future objects on the same ids to have invalid data.
Propagate the errno values from syscalls and use them in the warning messages to give info about what is going on.
don't leave the port and node names blank but fill in some defaults that include the node/port id.
We would timestamp within an unlikely block, which would introduce additional jitter to current_time, which would have an impact on the performance of the timer sensitive code.
Because it's easy to do and avoids some runtime checks.
Make a rtprio-server and rtprio-client option. Leave the server priority by default to 88 but lower client priority to 83. JACK does something similar by setting clients to rtprio-server - 5. Make module-rt use the client priority by default and bump the server priority explicitly in the config file. Leave the pulse-server to the default rtprio-client, there is no reason to lower this any further because it is really just a regular client. Bump the ffado packetizer thread to rtprio-server + 5 because that is also what JACK does. 88 is still much higher than the value of 60 that JACK uses in Fedora but now this is at least configurable.
Make a socket of the same address family as the address we're going to bind it to. When the source.ip is not specified, use the default in the same address family as the sap.ip. Probe the interface for addresses of the same family as the sap.ip. Makes pactl load-module module-rtp-recv sap_address=:: work. See #3851
Use the default source.ip from the same address family as the destination.ip. See #3851
Update doc/ from master branch. tutorial: fix s16 scale and add some docs doc: add 'Configuration' page doc: disable deprecated list doc: fix some doxygen warnings doc: put new pulse modules to right place doc: filter some constructs that confuse doxygen doc: Fix typo 'statis' -> 'static' doc: include pipewire-pulse modules explanations also on man page doc: add pw-v4l2.1 and spa-*.1 doc: add pw-reserve.1 doc: internals/access: update documentation vs current state
If the originating source is IPv6, the A bit is set and the source is 16 bytes compared to 4 bytes for IPv4. Fixes #3851
We need to load and apply the overrides in the order: If absolute config path, use only that. If environment variable, use only that. Else /usr/share/pipewire/*.conf.d/ /etc/pipewire/*.conf.d/ $HOME/.config/pipewire/*.conf.d/ Before this patch we would first apply $HOME and then /etc and /usr, which is not expected.
My making the phase into a float, the resampler can do finer grained adjustments, which should improve the stability of adaptive resampling
The rate we get from dlls can have a subsample precision. However, the check for using process_copy is in sample precision. This means that an adaptive stream will oscillate rather then lock into the exact rate.
All pw_core event handlers (`on_core_*()`) currently receive a pointer to `struct data`, not `struct proxy_data`; as can be seen from the `pw_core_add_listener()` call in `main()`. Fixes: cacdcc1 ("pw-mon: add filter param to hide props and/or params") Fixes #3997
Using `g_autoptr` and related modern helpers more often makes the code smaller, easier to follow and maintain. No behavior changes intended. (cherry picked from commit 594e3fa)
Some clients like many camera apps, including Cheese or Snapshot, trigger a lot of unnessecary renegotiations. While arguably that should be solved on a Gstreamer level, we can help out by checking if the preferred new caps are the same that are already in use and skip the renegotiation in this case. This allows several apps to e.g. take pictures without a slow and heavy stream restart. (cherry picked from commit e2e8cf7)
gst_pad_get_current_caps may return NULL and passing that into gst_caps_is_equal may result in fatal critical log due to the "g_return_val_if_fail (GST_IS_CAPS (caps1)" check. Fix by checking for NULL to avoid this. (cherry picked from commit 95127d8)
Check for NULL when importing a buffer and log a message instead of trying to deref the NULL pointer and crash. Add some more logging to mem when importing a bad fd. See #3998
Surrogate escapes must add not or 0x10000, as the specified bits go up 0xfffff.
`object::type` was not cleared in the error path. Fix that by calling `object_destroy()`, which takes care of it.
`registry_event_global()` creates an `object` object for every object, not just those matching `data::pattern`. However, previously `registry_event_global_remove()` only destroyed those objects that matched the given pattern. Fix that by destroying every object. Fixes #4001 Fixes 47e1f38 ("pw-dump: also dump object removal")
Libcamera formats are generally little-endian, matching DMA DRM fourccs, while PW ones are big-endian. Thus we have to invert the order. Only RGB and BGR where tested, as these are the formats currently supported by the software ISP. This fixes inverted red and blue in Snapshot on the Librem5 and Pinephone (OG). See also gstlibcamera-utils.cpp in libcamera.
Clear queued buffers when stopping consuming, to ensure that all buffers are usable when we start again. Do for A2DP as we already do for SCO.
We delay the audio a bit to keep packet intervals equal, which keeps some data in buffers. In theory the calculation keeps one buffer free, but it doesn't explicitly keep "extra" buffer space so in theory might flush too late and next process() might not have free buffers. However, as we encode next packet right away this shouldn't really occur... Try to keep one extra spare buffer free so that the flush time is certainly early enough.
The atomic operations were used in an attempt to limit the amount of wakeups done on the eventfd. It's however racy and causes trouble in some cases. Remove the atomic operations all together, it's probably not worth optimizing this too much.
The larger events need to be copied into the target buffer at the same offset as the source buffer or else we overwrite the header and make a corrupt buffer.
Avoid trying to remove a port before it has been added.
Add macro SPA_CMP to do 3-way comparisons safely, and use it to avoid signed integer overflows. Fix also float/double comparisons (previously 0.1 == 0.8 since cast to return type int). Fix Id/Bool comparisons so they can return negative value.
Handle allocation failure and size overflow.
It seems to be required to unselect it as default node in wireplumber. Fixes #4023
|
Isn't this a bit "old" than 1.2.1? |
|
Wireplumber requires a newer version of 1.0.X to build for Jammy. Meanwhile, 1.2.X has regressions which might be solved by a newer version of Wireplumber. |
jacobgkau
requested changes
Jul 18, 2024
There was a problem hiding this comment.
This one's working fine on pang12.
This does seem to break GNOME Shell screen recording more than it already was. There have already been some known issues with that, such as the cursor not showing up properly in the recording, and the recording only working once per GNOME Shell session unless the gjs process is manually killed. However, with this update, the recordings basically stop working entirely; they only seem to capture one or two frames, and produce a video that only lasts a split second (when played via MPV or VLC).
I've confirmed this regression occurs on multiple machines, on Intel, AMD, and NVIDIA graphics. The screen recorder worked on all three machines prior to the update, and no longer works after the update. I did notice that the filesize for the broken recordings are still comparable to the working ones. Based on the behavior in different video players, it seems likely this is some kind of reported framerate or other encoding error besides simply not saving the data.
The rest of regression testing passed. I could see a case for giving up on the built-in screen recording since it already worked poorly, but on the other hand, there has been a workaround available to at least make it work, and we know from issues such as pop-os/gnome-shell#134 that people are still using the functionality. For that reason, I think that should count as a blocker from a QA standpoint.
Full checklist:
- Built-in speakers and microphone still work, volume adjustment still plays sound.
- GNOME Sound Recorder still records and plays back.
- Audacity records and plays back.
- JACK mode
- ALSA mode
- DaVinci Resolve audio still works as expected (without lag) & shows up in pavucontrol.
- Volume adjustment sounds stop working when Resolve is open, not a regression.
- Ardour works in JACK mode, Bitwig works in native PipeWire mode.
- Output via GPU (e.g. HDMI and DisplayPort audio) works.
- Intel GPU system
- NVIDIA GPU system
- AMD GPU system
- Headphone/microphone and combo headset both work.
- Built-in GNOME Shell screen recording still works.
- Option to show cursor is broken, stuck in top-left in recording (not a regression).
- Only starts recording once per GNOME Shell restart (not a regression).
- Only records for a brief instant (this is a regression).
- Still able to share screen, webcam, and microphone via Jitsi in Firefox.
- Bluetooth microphone input works with no stuttering.
|
any progress on this? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Required by pop-os/wireplumber#19