Do not add failed push operations in the BufferedHistory #1323
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
peaBerberian
added
the
bug
peaBerberian
force-pushed
the
fix/gc-quotaexceedederror-skip-v3
branch
from
cd75250
to
f5d78c8
Compare
src/core/segment_buffers/implementations/audio_video/audio_video_segment_buffer.ts
Outdated
Show resolved
Hide resolved
peaBerberian
force-pushed
the
fix/gc-quotaexceedederror-skip-v3
branch
from
f5d78c8
to
59aa510
Compare
Florent-Bouisset
approved these changes
Dec 11, 2023
This commit fixes a kind of rare issue that is tricky to reproduce that
may happen after a `QuotaExceededError` is encountered when pushing a
chunk to the audio and video `SourceBuffer` due to memory limitations.
In that situation, we could be left in the following scenario:
1. We're loading a chunk and pushing it to the `SourceBuffer`
2. The `SourceBuffer` throws a `QuotaExceededError`, generally this
means that the device needs more memory to add that chunk
3. The SegmentBuffer code still adds the chunk to its `SegmentInventory`
(as it may still have been partially pushed to the SourceBuffer,
we're not 100% sure of what happened on the lower-level code) and
to that `SegmentInventory`'s `BufferedHistory`.
4. The stream code catches the error, try to remove some buffered data
and retry to push the same chunk.
5. Same QuotaExceededError
6. The SegmentBuffer code agains, re do the whole `SegmentInventory` +
`BufferedHistory` insertion thing for the same reasons
7. The stream code re-catches the issue, here it takes more drastic
measures, mainly reducing the amount of buffer that is built in
advance.
8. The stream code then check if it re-needs the chunk. However, it
sees through the `BufferedHistory` that it already tried to load it
two times, and that in those two times, the browser did not buffer
it (there ignoring that there was a `QuotaExceededError` involved).
9. The stream code thus assumes a browser/device issue with that
segment, and loads the next one instead.
10. When it's time to decode the skipped chunk, the RxPlayer
automatically seek over it, which is good, but we would have
prefered to actually log it.
This commit fixes the issue by clearly identifying in the
`SegmentInventory` chunks whose push operation had an issue. Those are:
1. never added to the `BufferedHistory`
2. always re-loaded if they become needed again
We could also have not added them to the `SegmentInventory` altogether
but I still think it's better to do it, as in the very rare chance that
chunk still had been at least partially pushed, our logic would be more
accurate.
peaBerberian
force-pushed
the
fix/gc-quotaexceedederror-skip-v3
branch
from
59aa510
to
79ebbfc
Compare
Merged
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.
This PR fixes a kind of rare issue that is tricky to reproduce that may happen after a
QuotaExceededErroris encountered when pushing a chunk to the audio and videoSourceBufferdue to memory limitations.In that situation, we could be left in the following scenario:
SourceBufferSourceBufferthrows aQuotaExceededError, generally this means that the device needs more memory to add that chunkSegmentInventory(as it may still have been partially pushed to the SourceBuffer, we're not 100% sure of what happened on the lower-level code) and to thatSegmentInventory'sBufferedHistory.SegmentInventory+BufferedHistoryinsertion thing for the same reasonsBufferedHistorythat it already tried to load chunk(s) from it two times, and that in those two times, the browser did not buffer it (there ignoring that there was aQuotaExceededErrorinvolved).This PR fixes the issue by clearly identifying in the
SegmentInventorychunks whose push operation had an issue. Those are:BufferedHistoryWe could also have not added them to the
SegmentInventoryaltogether but I still think it's better to do it, as in the very rare chance that chunk still had been at least partially pushed, our logic would be more accurate.