Break infinite loop for corrupted radio stream#422
Closed
ljufa wants to merge 152 commits intopdeljanov:rel-0.5from
Closed
Break infinite loop for corrupted radio stream#422ljufa wants to merge 152 commits intopdeljanov:rel-0.5from
ljufa wants to merge 152 commits intopdeljanov:rel-0.5from
Conversation
The symphonia-format-wav crate is superseded by symphonia-format-riff.
The next_packet call now returns Result<Option<Packet>> instead of Result<Packet>. A return value of Ok(None) indicates when the media ends. Whereas an UnexpectedEof error now always indicates an unexpected end to the media. Fixes pdeljanov#246, pdeljanov#134.
* Scoring The score function is now provided a scoped stream instead of a buffer. * Preference Tiers Allow format and metadata readers registered in Probe to be registered with a preference tier. There are three preference tiers: preferred, standard, and fallback. Symphonia will always register its first-party implementations of readers in the standard tier. Users that prefer to use their own implementations over Symphonia's may register them at the preferred tier. Likewise, if a user would prefer a first-party Symphonia implementation over their own implementation, they may register their implementation in the fallback tier. This capability is particularly important when different implementations may only partially support a format. * Probe Options A probe may now be instantiated with a set of options. The maximum number of bytes probe may scan from the source is now configurable. The maximum number of bytes any particular reader may consume while scoring may also be limited.
Instead, an initial set of metadata may be provided to a format reader upon creation. The reader is expected to use this metadata as the start of its own metadata log. This allows users to access all metadata via. the reader instead of handling metadata found through probing seperately.
Add an explicit lifetime to the `MediaSource` to support non-static media sources. Fixes pdeljanov#117, pdeljanov#171.
Since the fft module now supports SIMD via. RustFFT, MDCT no longer needs a SIMD and non-SIMD implementation.
This crate is intended to contain Symphonia first-party codec and format support code shared by multiple format readers and decoders.
Author
|
Atm, I could not find any stream that fails immediatelly and consistently, it happens randomly (might be after 1h). |
This is to prepare for adding decode timestamps (DTS) for video support.
Signed timestamps better support gapless playback of audio and are required for video decode timestamps. The following new standards are set out for format readers: - The number of frames and duration of a Track always exclude delay and padding frames. - The PTS, DTS, and duration of a Packet should be as written by the container. If the timing information must be synthesized by the reader, then negative PTS should be used for delay frames. The duration should exclude both delay and padding frames. Encoder delay and padding to be trimmed should be signalled via the trim_start and trim_end fields. The sum of duration and trim fields equal the duration decoded. For audio tracks, the DTS should equal the PTS. - Seeking to delay or padding PTS should be supported. - Decoders are responsible for trimming the output if gapless playback is enabled. Timestamps and durations are now represented by newtype wrappers of i64 and u64, respectively. The wrappers do not implement arithmetic operations that may panic or wrap. Rather, checked arithmetic operations are provided instead. Since i64 is not capable of representing a full unsigned 64-bit range (as supported by many media containers), these interfaces ensure overflows are always considered by the developer. The following guidelines are put forward for handling timestamp overflow: - Timestamp overflow should be lazily handled only once that timestamp is reached during playback or requested in a seek. - When timestamp overflow occurs, a warning should be logged. - A FormatReader::next_packet implementation should return Ok(None) since the stream has effectively ended. - A FormatReader::seek implementation should return an out-of-range seek error since the stream effectively ended before the overflow. - A MetadataReader should ignore unrepresentable timestamps (e.g., ignore a chapter whose start time overflows). The guiding principle for timestamp overflow is to support playing as much of the media as possible before bailing out. Given the range of an i64, this should never realistically happen in real-world media streams.
PacketBuilder uses the typestate builder pattern to enforce the correct construction of a valid packet.
It is a common type shared between format readers and decoders so it should be promoted out of the formats module.
This commit adds validation checks to the FLAC decoder to prevent potential panics or invalid behavior when processing malformed streams. - Verify frame channel count matches the stream information. - Ensure dropped bits per sample does not exceed frame bits per sample. - Validate that the predictor order does not exceed the block size.
…eljanov#442) Expands the fuzzing infrastructure of Symphonia and include fixes for critical bugs found in the FLAC and ALAC decoders during initial fuzzing. Fuzzer Changes: * Added component-level fuzzers for AAC, ADPCM, ALAC, FLAC, MPA, PCM, and Vorbis decoders. * Added a helper script to easily run all fuzzers. ALAC Fixes: * Limit the ALAC magic cookie maximum frame length to 65,536 frames to prevent malicious headers from causing excessive memory allocation (OOM). * Add validation to ensure partial frames does not exceed the maximum buffer size. * Add bounds checks to handle empty or short output buffers gracefully, preventing index out of bounds panics. FLAC fixes: * Disallow bits per sample > 32 from codec parameters as that is not supported by FLAC. * Disallow FLAC frames with a block size > the output buffer's length. Fixes pdeljanov#149.
Check the fuzzers as part of the CI workflow to ensure they remain buildable. Improvement on pdeljanov#149.
This PR expands the fuzzing infrastructure for Symphonia to include new fuzzing target for each type of supported demuxer. Continued work on pdeljanov#149
Add a simple builder to each to make construction easier.
Author
|
I found one url that fails and can be used to reproduce this issue |
Reduce the binary size between about 60kB (with opt-simd) to 600kB ( without opt-simd) by boxing some lookup tables generated at runtime with lazy_static. Despite their runtime initialization, their uninitialized state does not appear to be located in the bss section of the binary. Instead, their uninitialized states are present in the data section, bloating the binary. This change caused a minor performance regression with the built-in (no SIMD) FFT implementation. Switched to a breadth-first FFT in an attempt to amortize the FFT twiddle factor lookup to as few times as possible. Ended up having a much greater performance uplift overall. Addresses pdeljanov#456. Co-authored-by: Jordan Bayles <jophba@chromium.org>
…deljanov#463) Xiph lacing is sometimes used by WebM and Matroska containers to pack Vorbis extradata in the CodecPrivate data. Fixes pdeljanov#460.
# Conflicts: # .github/workflows/ci.yml # symphonia-bundle-flac/Cargo.toml # symphonia-bundle-flac/src/demuxer.rs # symphonia-bundle-mp3/Cargo.toml # symphonia-bundle-mp3/src/demuxer.rs # symphonia-check/Cargo.toml # symphonia-codec-aac/Cargo.toml # symphonia-codec-aac/src/aac/mod.rs # symphonia-codec-aac/src/adts.rs # symphonia-codec-adpcm/Cargo.toml # symphonia-codec-alac/Cargo.toml # symphonia-codec-opus/Cargo.toml # symphonia-codec-pcm/Cargo.toml # symphonia-codec-vorbis/Cargo.toml # symphonia-codec-vorbis/src/codebook.rs # symphonia-codec-wavpack/Cargo.toml # symphonia-common/LICENSE # symphonia-core/Cargo.toml # symphonia-core/src/audio.rs # symphonia-core/src/codecs.rs # symphonia-core/src/formats.rs # symphonia-core/src/probe.rs # symphonia-format-caf/Cargo.toml # symphonia-format-isomp4/Cargo.toml # symphonia-format-isomp4/src/atoms/esds.rs # symphonia-format-isomp4/src/atoms/flac.rs # symphonia-format-isomp4/src/atoms/ilst.rs # symphonia-format-isomp4/src/atoms/mod.rs # symphonia-format-isomp4/src/atoms/opus.rs # symphonia-format-isomp4/src/atoms/stsd.rs # symphonia-format-isomp4/src/demuxer.rs # symphonia-format-mkv/Cargo.toml # symphonia-format-mkv/src/demuxer.rs # symphonia-format-mkv/src/ebml.rs # symphonia-format-ogg/Cargo.toml # symphonia-format-ogg/src/logical.rs # symphonia-format-riff/Cargo.toml # symphonia-format-riff/src/aiff/chunks.rs # symphonia-format-wav/Cargo.toml # symphonia-metadata/Cargo.toml # symphonia-play/Cargo.toml # symphonia-utils-xiph/Cargo.toml # symphonia/Cargo.toml # symphonia/src/lib.rs
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
10 participants
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.
In my https://github.com/ljufa/rsplayer that I am using and testing daily, I've experienced infinite loop created by some online radio streams. (I can provide some examples if needed)
I guess this fix is not generic enough to solve corupted stream problems among all streams/codecs but in my case it helped. Open for ideas to solve it in more elegant way...