mediadecode_ffmpeg/resampler.rs
1//! [`mediadecode::resampler::AudioResampler`] impl backed by
2//! `libswresample`.
3//!
4//! Converts rate, sample format and channel layout between two specs
5//! fixed at construction — [`FfmpegResampler::new`] takes both, because
6//! neither end is discoverable and neither is a constant. The source is
7//! whatever the file holds; the target is whatever the consumer wants,
8//! and consumers disagree (16 kHz mono for a speech model, 48 kHz for
9//! an audio-event one, from the same track at the same time).
10//!
11//! # Output timestamps
12//!
13//! `swr` is a delay line: it needs future input to produce present
14//! output, so at any moment a filter's worth of samples is inside it.
15//! Timestamps are therefore *counted*, not computed per call — the
16//! output timeline is anchored on the first input timestamp and
17//! advanced by the number of samples actually produced. The frames
18//! drained after EOF continue that same line rather than restarting it,
19//! and no arithmetic anywhere depends on how many samples a given
20//! `swr_convert_frame` happened to yield.
21
22use std::{
23 collections::VecDeque,
24 ptr::{addr_of, read_unaligned},
25};
26
27use derive_more::{IsVariant, TryUnwrap, Unwrap};
28use ffmpeg_next::{
29 ChannelLayout,
30 codec::Parameters,
31 ffi::{
32 AV_NOPTS_VALUE, AVChannelOrder, AVMatrixEncoding, AVSampleFormat, av_channel_layout_from_mask,
33 av_frame_get_buffer, swr_build_matrix2,
34 },
35 format::Sample,
36 frame,
37 software::resampling,
38};
39use mediadecode::{
40 Received, Sent, Timebase, Timestamp,
41 frame::{AudioFrame, Plane},
42 resampler::AudioResampler,
43};
44use mediaframe::audio::ChannelLayoutDescription;
45
46use crate::{
47 Error, Ffmpeg, extras::AudioFrameExtra, limits::FrameLimits, sample_format::SampleFormat,
48};
49
50/// The frame type a resampler accepts and produces, on lane `C`.
51///
52/// Written as a projection rather than a bounded alias so the bound
53/// lives on the items that need it — `type_alias_bounds` is not
54/// enforced anyway, and a bound written where it is not enforced reads
55/// like a promise the compiler is keeping.
56type Frame<C> = AudioFrame<
57 SampleFormat,
58 ChannelLayoutDescription,
59 AudioFrameExtra,
60 <C as crate::FfmpegCarrier>::Buffer,
61>;
62
63/// One end of a conversion: sample rate, sample format, channel layout.
64///
65/// Spelled in FFmpeg's own vocabulary because construction is off the
66/// [`AudioResampler`] trait and this is the backend that has to be
67/// handed to `swr_alloc_set_opts2`. [`FfmpegResampler`] restates the
68/// source spec in the vocabulary a decoded frame carries, so the
69/// mid-stream check compares like with like without the caller ever
70/// seeing two dialects.
71#[derive(Copy, Clone, Debug, PartialEq, Eq)]
72pub struct ResampleSpec {
73 rate: u32,
74 format: Sample,
75 layout: ChannelLayout,
76}
77
78impl ResampleSpec {
79 /// Constructs a spec from its three parts.
80 ///
81 /// Deliberately total and `const`: a spec is a description, and
82 /// describing something `swr` cannot convert is not itself an error.
83 /// [`FfmpegResampler::new`] is the choke point every construction
84 /// route passes through, and it is what refuses a rate, a format or a
85 /// channel layout this backend cannot honour — see
86 /// [`FfmpegResampler::new`] for the roster and
87 /// [`ResampleError::UnsupportedLayout`] for why a layout can be
88 /// refused at all.
89 #[inline]
90 pub const fn new(rate: u32, format: Sample, layout: ChannelLayout) -> Self {
91 Self {
92 rate,
93 format,
94 layout,
95 }
96 }
97
98 /// The spec a track *declares*, read off the codec parameters a
99 /// [`crate::FfmpegDemuxer`] track row carries
100 /// (`track.extra().parameters()`) — the "source from `TrackInfo`"
101 /// path.
102 ///
103 /// Returns `None` for a non-audio track, for one whose declared
104 /// sample format is `AV_SAMPLE_FMT_NONE` (a codec whose format is
105 /// only known once its decoder opens), and for a custom or ambisonic
106 /// channel layout — see [`Self::from_decoder`] for the first case and
107 /// the note on [`unspecified_layout`] for the last.
108 pub fn from_parameters(parameters: &Parameters) -> Option<Self> {
109 // Before `medium()`, which dereferences the pointer inside
110 // ffmpeg-next. `Parameters`' safe constructors hand back a
111 // null-backed value when FFmpeg's allocation failed and report
112 // nothing, so a caller can arrive here holding one without ever
113 // having been told. Parameters that were never allocated describe
114 // no audio, which this function already has a word for.
115 // SAFETY: reading the pointer without dereferencing it.
116 if unsafe { parameters.as_ptr() }.is_null() {
117 return None;
118 }
119 if !crate::boundary::media_kind_of(parameters).is_audio() {
120 return None;
121 }
122 // SAFETY: `parameters` keeps the `AVCodecParameters` live; every
123 // read below goes through the raw pointer and none of them
124 // materialises a bindgen enum out of foreign memory.
125 let par = unsafe { parameters.as_ptr() };
126 let rate = unsafe { (*par).sample_rate }.max(0) as u32;
127 if rate == 0 {
128 return None;
129 }
130 let format = SampleFormat::from_raw(unsafe { (*par).format }).to_ffmpeg()?;
131 let layout = unsafe { layout_from_raw(addr_of!((*par).ch_layout)) }?;
132 Some(Self::new(rate, format, layout))
133 }
134
135 /// The spec an opened decoder will actually produce — its rate,
136 /// sample format and channel layout, straight off the codec context.
137 ///
138 /// Reach it through
139 /// [`FfmpegAudioStreamDecoder::inner`](crate::FfmpegAudioStreamDecoder::inner).
140 /// `None` on a custom or ambisonic layout, and on a context whose
141 /// sample format is still unset (a decoder that has not been opened).
142 pub fn from_decoder(decoder: &ffmpeg_next::decoder::Audio) -> Option<Self> {
143 // SAFETY: `decoder` keeps the `AVCodecContext` live. `sample_fmt`
144 // is read as the raw integer it is rather than through
145 // `decoder.format()`, which would construct an `AVSampleFormat`
146 // out of foreign memory.
147 let ctx = unsafe { decoder.as_ptr() };
148 // Same reason as `from_parameters`: `codec::Context::new()` is a
149 // safe constructor over an unchecked `avcodec_alloc_context3`, so a
150 // decoder can be null-backed without anyone having been told.
151 if ctx.is_null() {
152 return None;
153 }
154 let format =
155 SampleFormat::from_raw(unsafe { read_unaligned(addr_of!((*ctx).sample_fmt).cast::<i32>()) })
156 .to_ffmpeg()?;
157 let rate = unsafe { (*ctx).sample_rate }.max(0) as u32;
158 if rate == 0 {
159 return None;
160 }
161 let layout = unsafe { layout_from_raw(addr_of!((*ctx).ch_layout)) }?;
162 Some(Self::new(rate, format, layout))
163 }
164
165 /// A layout that names a channel *count* and nothing else —
166 /// `AV_CHANNEL_ORDER_UNSPEC`.
167 ///
168 /// Not a degenerate case: a WAV file without a `WAVE_FORMAT_EXTENSIBLE`
169 /// channel mask genuinely declares no layout, and FFmpeg faithfully
170 /// reports it as unspecified in the codec parameters, in the codec
171 /// context, and on every decoded frame. Substituting a default layout
172 /// would make the source spec disagree with the frames it is supposed
173 /// to describe, and every `send_frame` would be refused as a
174 /// mid-stream change. `swr` accepts an unspecified layout at either
175 /// end and maps the channels positionally.
176 #[inline]
177 pub fn unspecified_layout(channels: i32) -> ChannelLayout {
178 // SAFETY: a zeroed `AVChannelLayout` is a valid value — `order`
179 // reads as `AV_CHANNEL_ORDER_UNSPEC`, the zero discriminant, and
180 // the union is documented as unused for that order.
181 unsafe {
182 let mut layout: ffmpeg_next::ffi::AVChannelLayout = std::mem::zeroed();
183 layout.nb_channels = channels.max(0);
184 ChannelLayout(layout)
185 }
186 }
187
188 /// Sample rate in Hz.
189 #[inline]
190 pub const fn rate(&self) -> u32 {
191 self.rate
192 }
193 /// Sample format.
194 #[inline]
195 pub const fn format(&self) -> Sample {
196 self.format
197 }
198 /// Channel layout.
199 #[inline]
200 pub const fn layout(&self) -> ChannelLayout {
201 self.layout
202 }
203 /// Channel count, from the layout.
204 #[inline]
205 pub fn channels(&self) -> i32 {
206 self.layout.channels()
207 }
208
209 /// The timebase output frames carry — one tick per output sample.
210 fn timebase(&self) -> Timebase {
211 Timebase::new(
212 1,
213 std::num::NonZeroI32::new(self.rate.min(i32::MAX as u32) as i32).unwrap_or(
214 // A zero-rate spec never reaches here: `new` is the only way in
215 // and every caller of it names a real rate. Falling back to
216 // one tick per second keeps the arithmetic total rather than
217 // panicking on a value that cannot occur.
218 std::num::NonZeroI32::new(1).expect("1 is non-zero"),
219 ),
220 )
221 }
222}
223
224/// `mediadecode::resampler::AudioResampler` impl wrapping
225/// `swresample`.
226///
227/// Construction is [`Self::new`], off the trait, taking both specs —
228/// see the trait's own documentation for why the target can never be a
229/// constant.
230pub struct CarrierResampler<C: crate::FfmpegCarrier> {
231 ctx: resampling::Context,
232 source: ResampleSpec,
233 target: ResampleSpec,
234 /// The source spec restated in the vocabulary a decoded `AudioFrame`
235 /// carries. The mid-stream check compares against these, not against
236 /// FFmpeg's dialect, so it never has to translate a frame.
237 source_format: SampleFormat,
238 source_layout: ChannelLayoutDescription,
239 /// The target spec in the vocabulary an output frame carries,
240 /// computed once at construction. Assembling a converted frame after
241 /// `swr` has run must not have to ask FFmpeg anything, because asking
242 /// can fail — see [`FfmpegResampler::prepare_output`].
243 target_format: SampleFormat,
244 target_layout: ChannelLayoutDescription,
245 /// The layouts `swr` is really configured with — see
246 /// [`initialized_layout`]. Every `AVFrame` this type stages or
247 /// allocates carries these, not the declared ones.
248 staged_source_layout: ChannelLayout,
249 staged_target_layout: ChannelLayout,
250 target_timebase: Timebase,
251 ready: VecDeque<Frame<C>>,
252 /// Next output timestamp, in target-rate ticks. `None` until the
253 /// first input frame anchors it.
254 next_pts: Option<i64>,
255 eof: bool,
256 /// `true` once the post-EOF tail has been drained to its end.
257 ///
258 /// **The terminal answer has to be terminal.** Without this latch,
259 /// every poll past the end re-enters the flush road — allocating an
260 /// output frame and asking `swr` again — because `swr_get_delay` can
261 /// keep reporting residual samples that `swr_convert_frame` will
262 /// never emit (observed: 16 samples left standing after the tail is
263 /// genuinely exhausted). That is harmless while allocation succeeds
264 /// and wrong when it does not: a session that has already answered
265 /// `Ended` would start answering with an allocation error instead,
266 /// turning a settled protocol state back into a fault. The latch
267 /// makes the end cheap and unconditional. `flush` clears it with the
268 /// rest of the session.
269 drained: bool,
270 /// What one converted frame may cost. See
271 /// [`Self::check_output_bytes`] for why a resampler needs a ceiling
272 /// of its own even when its input already had one.
273 limits: FrameLimits,
274 /// The lane. Zero-sized: it selects how a produced plane is carried
275 /// — shared out of the output `AVFrame` or copied out of it — and
276 /// nothing else about the conversion.
277 _carrier: core::marker::PhantomData<C>,
278}
279
280impl<C: crate::FfmpegCarrier + crate::CarrierOps> CarrierResampler<C> {
281 /// Opens a resampler between two explicit specs.
282 ///
283 /// Both are required and neither is inferred. The source is what the
284 /// decoder will hand over — read it off the track
285 /// ([`ResampleSpec::from_parameters`]) or off the opened decoder
286 /// ([`ResampleSpec::from_decoder`]). The target is the caller's, and
287 /// is options: 16 kHz mono for a speech model, 48 kHz for an
288 /// audio-event one, both from the same track.
289 ///
290 /// # The choke point
291 ///
292 /// [`ResampleSpec::new`] is `const` and total, so this is where both
293 /// ends are checked — every construction route (`from_parameters`,
294 /// `from_decoder`, the public constructor) passes through here, and
295 /// nothing hazardous reaches `swr` or a staged `AVFrame` behind it:
296 ///
297 /// - a rate of zero, or one past `c_int`
298 /// ([`ResampleError::UnsupportedRate`]);
299 /// - `AV_SAMPLE_FMT_NONE` ([`ResampleError::UnsupportedFormat`]);
300 /// - a channel layout that is neither native nor unspecified, or one
301 /// naming no channels ([`ResampleError::UnsupportedLayout`]).
302 ///
303 /// `limits` bounds what one **converted** frame may cost.
304 ///
305 /// [`FrameLimits`] rather than a seat of this seam's own: the
306 /// quantity is bytes of one produced audio frame, which is exactly
307 /// what [`FrameLimits::max_frame_bytes`] means everywhere else in
308 /// this crate, and one number for "what a frame may cost" is worth
309 /// more than a second vocabulary. [`FrameLimits::max_pixels`] is
310 /// unused here, as it is on the audio decode path, and for the same
311 /// reason: audio has no pixels.
312 pub(crate) fn new_impl(
313 source: ResampleSpec,
314 target: ResampleSpec,
315 limits: FrameLimits,
316 ) -> Result<Self, ResampleError> {
317 check_spec(&source, SpecEnd::Source)?;
318 check_spec(&target, SpecEnd::Target)?;
319
320 // The layouts `swr` is really configured with, resolved *before*
321 // the pair is judged — because the conversion that will run is
322 // between these two, not between the two that were declared. An
323 // unspecified layout becomes FFmpeg's default for its channel count
324 // (twenty-four unspecified channels are 22.2), so judging the
325 // declared pair let exactly the routing the explicit 22.2 refusal
326 // blocks walk in through the unspecified door.
327 let staged_source_layout = initialized_layout(source.layout);
328 let staged_target_layout = initialized_layout(target.layout);
329 check_pair(&staged_source_layout, &staged_target_layout)?;
330 let ctx = open_context(&source, &target, staged_source_layout, staged_target_layout)?;
331
332 let source_format = SampleFormat::from_ffmpeg(source.format);
333 let target_format = SampleFormat::from_ffmpeg(target.format);
334 // SAFETY: the layout is a live `ChannelLayout` owned by this scope.
335 let target_layout =
336 crate::channel_layout::channel_layout_description_from_ffmpeg(&staged_target_layout);
337 // SAFETY: the layout is a live `ChannelLayout` owned by `source`
338 // for the duration of this call.
339 let source_layout =
340 crate::channel_layout::channel_layout_description_from_ffmpeg(&source.layout);
341 let target_timebase = target.timebase();
342
343 Ok(Self {
344 ctx,
345 source,
346 target,
347 source_format,
348 source_layout,
349 target_format,
350 target_layout,
351 staged_source_layout,
352 staged_target_layout,
353 target_timebase,
354 ready: VecDeque::new(),
355 next_pts: None,
356 eof: false,
357 drained: false,
358 limits,
359 _carrier: core::marker::PhantomData,
360 })
361 }
362
363 /// The spec frames must arrive in.
364 #[inline]
365 pub(crate) const fn source_impl(&self) -> &ResampleSpec {
366 &self.source
367 }
368
369 /// The spec frames leave in.
370 #[inline]
371 pub(crate) const fn target_impl(&self) -> &ResampleSpec {
372 &self.target
373 }
374
375 /// Borrows the wrapped `swr` context.
376 #[inline]
377 pub(crate) const fn inner_impl(&self) -> &resampling::Context {
378 &self.ctx
379 }
380
381 /// Samples still inside the delay line, counted at the output rate.
382 #[inline]
383 pub(crate) fn delay_impl(&self) -> i64 {
384 self.ctx.delay().map_or(0, |d| d.output.max(0))
385 }
386
387 /// Refuses a frame whose shape is not the source spec.
388 fn check_source(&self, frame: &Frame<C>) -> Result<(), ResampleError> {
389 if frame.sample_rate() != self.source.rate
390 || *frame.sample_format() != self.source_format
391 || *frame.channel_layout() != self.source_layout
392 {
393 return Err(ResampleError::SourceChanged(SourceChanged::new(
394 self.source.rate,
395 self.source_format,
396 frame.sample_rate(),
397 *frame.sample_format(),
398 )));
399 }
400 Ok(())
401 }
402
403 /// Where a frame's timestamp lands on the output timeline, or `None`
404 /// when it carries none.
405 ///
406 /// Rescaled with the **checked** rung, and before anything is staged.
407 /// `Timestamp::rescale_to` saturates, and both ends of that clamp are
408 /// wrong here: a positive one reaches the counted timeline's checked
409 /// addition only after `swr` has consumed the input, leaving a
410 /// session no caller can retry; a negative one lands on `i64::MIN`,
411 /// which *is* `AV_NOPTS_VALUE`, so the conversion back reads the
412 /// frame as carrying no timestamp at all and an extreme timestamp is
413 /// silently erased. A timestamp that does not fit the output timeline
414 /// is refused by name, with the resampler untouched.
415 fn anchor_of(&self, frame: &Frame<C>) -> Result<Option<i64>, ResampleError> {
416 let Some(timestamp) = frame.pts() else {
417 return Ok(None);
418 };
419 let ticks = timestamp.pts();
420 let out_of_range = || ResampleError::TimestampOutOfRange(TimestampOutOfRange::new(ticks));
421 // `AV_NOPTS_VALUE` is a sentinel, not a time. A frame carrying it
422 // as a value says something contradictory, and anchoring on it
423 // would produce output frames that report no timestamp.
424 if ticks == AV_NOPTS_VALUE {
425 return Err(out_of_range());
426 }
427 let rescaled = timestamp
428 .timebase()
429 .checked_rescale(ticks, self.target_timebase)
430 .ok_or_else(out_of_range)?;
431 if rescaled == AV_NOPTS_VALUE {
432 return Err(out_of_range());
433 }
434 Ok(Some(rescaled))
435 }
436
437 /// Stages a decoded frame as an `AVFrame` swr can read.
438 ///
439 /// Geometry is settled **before** anything is allocated. A frame's
440 /// header is a claim, not a fact: `nb_samples` comes from the same
441 /// foreign memory as the planes it describes, and sizing an
442 /// allocation off it first would let a forged frame with a
443 /// twelve-byte plane ask for tens of gigabytes on its way to being
444 /// refused.
445 fn stage_input(&self, frame: &Frame<C>) -> Result<frame::Audio, ResampleError> {
446 let samples = frame.nb_samples() as usize;
447 let channels = self.source.channels();
448
449 // What the *format* requires, not what the allocated frame reports
450 // — the frame does not exist yet.
451 let planes = if self.source.format.is_planar() {
452 // `check_spec` proved this positive at construction.
453 channels as usize
454 } else {
455 1
456 };
457 let found = frame.plane_count() as usize;
458 if planes > found {
459 return Err(ResampleError::PlaneCount(PlaneCount::new(planes, found)));
460 }
461 let bytes = plane_bytes(self.source.format, samples, channels)
462 .ok_or(ResampleError::SampleCount(SampleCount::new(samples)))?;
463 for plane in frame.planes().iter().take(planes) {
464 let src = plane.data_ref().as_ref();
465 if src.len() < bytes {
466 return Err(ResampleError::PlaneCount(PlaneCount::new(bytes, src.len())));
467 }
468 }
469
470 // Only now, with every plane proved long enough for the sample
471 // count that sizes this allocation.
472 let mut input = new_audio_frame(
473 self.source.format,
474 samples,
475 self.source.rate,
476 self.staged_source_layout,
477 )?;
478 // What the allocation really produced. `data_mut` panics past its
479 // own plane count, and this crate does not put a panic on a path
480 // that reads foreign geometry.
481 let staged = input.planes();
482 if staged < planes {
483 return Err(ResampleError::PlaneCount(PlaneCount::new(planes, staged)));
484 }
485 for (index, plane) in frame.planes().iter().take(planes).enumerate() {
486 let src = plane.data_ref().as_ref();
487 let dst = input.data_mut(index);
488 if dst.len() < bytes {
489 return Err(ResampleError::PlaneCount(PlaneCount::new(bytes, dst.len())));
490 }
491 dst[..bytes].copy_from_slice(&src[..bytes]);
492 }
493 Ok(input)
494 }
495
496 /// The most samples the next conversion could produce: the delay
497 /// line's contents plus `in_samples` of new input, rescaled to the
498 /// output rate and rounded up.
499 ///
500 /// Separate from the allocation because it is also the preflight the
501 /// output timeline is checked against — *before* `swr` consumes
502 /// anything, so a refusal leaves the session where a caller can retry
503 /// it.
504 fn output_capacity(&self, in_samples: i64) -> Result<usize, ResampleError> {
505 let delay_in = self.ctx.delay().map_or(0, |d| d.input.max(0));
506 let total = delay_in.saturating_add(in_samples).max(0) as i128;
507 let scaled = (total * i128::from(self.target.rate) + i128::from(self.source.rate) - 1)
508 / i128::from(self.source.rate).max(1);
509 // One extra sample of headroom: swr rounds its own accounting, and
510 // an output frame one short would silently push the remainder into
511 // the internal FIFO where the pts accounting cannot see it until
512 // the next call.
513 let samples = scaled + 1;
514 // `av_frame_get_buffer` takes the count as a `c_int`. A request
515 // past that is refused by name rather than clamped: a silently
516 // shortened output frame is a stream that loses samples.
517 if samples > i128::from(i32::MAX) {
518 // Saturating only for a count past `usize` itself, which no
519 // machine could hold either way.
520 return Err(ResampleError::SampleCount(SampleCount::new(
521 usize::try_from(samples).unwrap_or(usize::MAX),
522 )));
523 }
524 Ok(samples.max(1) as usize)
525 }
526
527 /// Refuses a conversion whose output would not fit the frame ceiling.
528 ///
529 /// # Why a resampler needs one even though its input had one
530 ///
531 /// [`Self::output_capacity`] bounds the output **sample count**, and
532 /// only at `i32::MAX` — the structural limit of
533 /// `av_frame_get_buffer`. Nothing in it bounds the *bytes*, and the
534 /// two are related by a ratio the caller does not control: the
535 /// capacity is `input_samples × target_rate / source_rate`, and a
536 /// source spec read off an untrusted container can say 1 Hz. One
537 /// second of 1 Hz mono input converted to 48 kHz stereo `f32` is
538 /// 384 KiB from 4 bytes — and the same input against a 1 Hz source
539 /// claim and a 192 kHz target is multi-gigabyte. The frame that
540 /// arrived was within *its* ceiling; the frame that leaves need not
541 /// be, so it gets its own judgement.
542 ///
543 /// Both allocations are covered: `av_frame_get_buffer`'s, and the
544 /// [`FfmpegBytes`] copy [`Self::finish_output`] makes from it.
545 fn check_output_bytes(&self, capacity: usize, channels: i32) -> Result<(), ResampleError> {
546 // **Priced as the allocator prices it, not as the samples weigh.**
547 // This used to multiply the tight plane length by the plane count,
548 // which is the payload's arithmetic and not
549 // `av_frame_get_buffer`'s: a one-sample eight-channel planar `s16`
550 // frame is 16 bytes of samples and a **768-byte** allocation,
551 // because every plane is aligned and padded on its own. A 16-byte
552 // ceiling admitted it.
553 //
554 // The overhead is under one percent on any frame big enough to
555 // care about, which is exactly why this went unseen — see
556 // [`crate::footprint`] for the measured table and for the rule the
557 // whole crate now keeps: a judge must dominate the allocator's
558 // arithmetic, not the payload's.
559 let bytes = crate::footprint::audio_frame_bytes(
560 ffmpeg_next::ffi::AVSampleFormat::from(self.target.format) as libc::c_int,
561 capacity,
562 channels.max(0) as usize,
563 )
564 .ok_or(ResampleError::OutputTooLarge(OutputTooLarge::new(
565 usize::MAX,
566 self.limits.max_frame_bytes(),
567 )))?;
568 if bytes > self.limits.max_frame_bytes() {
569 return Err(ResampleError::OutputTooLarge(OutputTooLarge::new(
570 bytes,
571 self.limits.max_frame_bytes(),
572 )));
573 }
574 Ok(())
575 }
576
577 /// Refuses a conversion whose output could not be labelled: the
578 /// timeline plus everything this call might produce has to stay
579 /// inside `i64`.
580 ///
581 /// Asked before `swr` sees a sample, like everything else that can
582 /// fail. [`Self::finish_output`] performs the same addition against
583 /// the count actually produced, which cannot exceed the capacity
584 /// checked here — so once this passes, that one cannot fail.
585 fn check_timeline(&self, anchor: Option<i64>, capacity: usize) -> Result<(), ResampleError> {
586 let pts = self.next_pts.or(anchor).unwrap_or(0);
587 let samples = capacity as i64;
588 if pts.checked_add(samples).is_none() {
589 return Err(ResampleError::TimestampOverflow(TimestampOverflow::new(
590 pts, samples,
591 )));
592 }
593 Ok(())
594 }
595
596 /// Allocates the output frame **and proves every plane the converted
597 /// frame will be read out of**, before `swr` is allowed to touch a
598 /// sample.
599 ///
600 /// This is the shape the whole seam is built around. Anything
601 /// fallible that runs *after* `swr_convert_frame` has consumed input
602 /// leaves a session no caller can act on: retrying feeds the same
603 /// samples twice, continuing loses them, and the delay line has moved
604 /// either way. The failure kept relocating — the timestamp addition,
605 /// the tail drain, then the output wrapping — so the fix is not
606 /// another check in another place but an ordering that leaves nothing
607 /// on the far side: the frame, every plane pointer it will be read
608 /// through, and the queue slot are all taken here, where failing
609 /// costs nothing but an error.
610 ///
611 /// **What 0.9 changed, and what it did not.** Through 0.8 this
612 /// function also *acquired* one `AVBufferRef` view per plane, because
613 /// wrapping a plane could fail and so had to happen on this side of
614 /// the conversion; [`Self::finish_output`] then narrowed each view to
615 /// what `swr` produced. The amputation removes the views — the output
616 /// planes are copied out afterwards instead — and with them the
617 /// failure that forced the acquisition to be early. What stays early
618 /// is the *proof*: every plane pointer is checked non-null and
619 /// checked to address `plane_len` bytes inside one of the frame's own
620 /// buffers here, so the copy on the far side has nothing left to
621 /// judge and [`Self::finish_output`] remains infallible.
622 fn prepare_output(&self, capacity: usize) -> Result<PreparedOutput<C>, ResampleError> {
623 let channels = self.target.channels();
624 let plane_count = if self.target.format.is_planar() {
625 // `check_spec` proved this positive at construction.
626 channels as usize
627 } else {
628 1
629 };
630 let plane_len = plane_bytes(self.target.format, capacity, channels)
631 .ok_or(ResampleError::SampleCount(SampleCount::new(capacity)))?;
632 // Linear in the sample count, which is what lets the post-run
633 // trim be a multiplication rather than another fallible call.
634 let per_sample = plane_bytes(self.target.format, 1, channels)
635 .ok_or(ResampleError::SampleCount(SampleCount::new(1)))?;
636
637 // **The byte ceiling, before `av_frame_get_buffer` and before the
638 // carrier copy that follows it.** Measured first, allocated second:
639 // the three quantities above are arithmetic over the target spec
640 // and cost nothing, so there is no reason for the frame to exist
641 // before the answer does.
642 self.check_output_bytes(capacity, channels)?;
643
644 let frame = new_audio_frame(
645 self.target.format,
646 capacity,
647 self.target.rate,
648 self.staged_target_layout,
649 )?;
650 if frame.planes() < plane_count {
651 return Err(ResampleError::PlaneCount(PlaneCount::new(
652 plane_count,
653 frame.planes(),
654 )));
655 }
656
657 let mut reserved: [Option<C::Reserved>; 8] = core::array::from_fn(|_| None);
658 for (index, slot) in reserved.iter_mut().enumerate().take(plane_count) {
659 // SAFETY: `frame` owns a live `AVFrame` this call just
660 // allocated; `data` is a public field and `plane_count` is
661 // within the eight slots `data` has.
662 let data_ptr = unsafe { (*frame.as_ptr()).data[index] };
663 if data_ptr.is_null() {
664 return Err(ResampleError::OutputBuffer(OutputBuffer::new(index)));
665 }
666 // SAFETY: the frame is live, and the helper only reads `buf[]`'s
667 // ranges to find the one containing `data_ptr`. Proving it here
668 // is what lets the carry on the far side of the conversion be
669 // unconditional — a copy on the owned lane, a view on the other,
670 // and neither has a proof left to make.
671 let backing =
672 unsafe { crate::convert::find_audio_backing_buffer(frame.as_ptr(), data_ptr, plane_len) }
673 .ok_or(ResampleError::OutputBuffer(OutputBuffer::new(index)))?;
674 // The carrier's claim on the plane, taken **here** — before `swr`
675 // consumes anything. See [`FfmpegCarrier::reserve`].
676 //
677 // SAFETY: `backing` is a live buffer of this frame's, proved just
678 // above to cover `plane_len` bytes from `data_ptr`; the frame is
679 // moved into `PreparedOutput` below and so outlives the commit.
680 let offset = unsafe { (data_ptr as usize).wrapping_sub((*backing).data as usize) };
681 // SAFETY: as above — the extent lies inside `backing`.
682 *slot = Some(
683 unsafe { C::reserve(backing, offset, plane_len) }
684 .ok_or(ResampleError::OutputBuffer(OutputBuffer::new(index)))?,
685 );
686 }
687
688 Ok(PreparedOutput {
689 frame,
690 reserved,
691 plane_count,
692 plane_len,
693 per_sample,
694 })
695 }
696
697 /// Turns a converted frame into a `mediadecode` one. **Infallible**,
698 /// by construction: every check it could have made was made in
699 /// [`Self::prepare_output`], and everything left here is arithmetic
700 /// over values this type owns plus a copy that cannot be refused.
701 ///
702 /// `None` when the conversion produced nothing — the delay line
703 /// swallowed the input, which is ordinary and not a failure.
704 fn finish_output(&mut self, mut prepared: PreparedOutput<C>) -> Option<Frame<C>> {
705 let produced = prepared.frame.samples();
706 if produced == 0 {
707 return None;
708 }
709 let pts = self.next_pts.unwrap_or(0);
710 // `check_timeline` ran before `swr` did, against a capacity that is
711 // never smaller than what came out, so this addition cannot leave
712 // `i64`. It is stated rather than checked because a check here
713 // would be an error path on the wrong side of the conversion —
714 // exactly what this design exists to remove.
715 debug_assert!(
716 pts.checked_add(produced as i64).is_some(),
717 "the timeline was preflighted against a capacity >= produced",
718 );
719 self.next_pts = Some(pts.saturating_add(produced as i64));
720
721 let bytes = prepared
722 .per_sample
723 .saturating_mul(produced)
724 .min(prepared.plane_len);
725 let plane_count = prepared.plane_count;
726 let mut planes: [Plane<C::Buffer>; 8] = core::array::from_fn(|_| Plane::new(C::empty(), 0));
727 for (index, slot) in planes.iter_mut().enumerate().take(plane_count) {
728 // Present for every index below `plane_count`: `prepare_output`
729 // fills exactly that many and returns an error otherwise.
730 let Some(reserved) = prepared.reserved[index].take() else {
731 continue;
732 };
733 // **The valid prefix, not the plane.** `plane_len` is what
734 // capacity was allocated for; `bytes` is what `swr` produced.
735 // Both lanes stop at the latter, for the same reason the decode
736 // road does: the tail is allocator memory nothing wrote, and a
737 // carrier's span is a span a consumer may read.
738 //
739 // SAFETY: the reservation covers `plane_len` bytes inside one of
740 // `frame`'s own buffers, and `bytes <= plane_len` by the `min`
741 // above. `swr_convert_frame` has written those bytes and does not
742 // replace the buffers; `frame` has been owned by `prepared` — and
743 // so kept alive — across the whole conversion. A view committed
744 // here outlives `prepared` by refcount, and the writing is over
745 // before the sharing begins: this resampler allocates a fresh
746 // output frame per conversion, so nothing ever writes into a
747 // buffer a delivered frame is reading.
748 *slot = Plane::new(unsafe { C::commit(reserved, bytes) }, bytes as u32);
749 }
750
751 Some(
752 AudioFrame::new(
753 self.target.rate,
754 produced as u32,
755 // Exact, not clipped: `check_spec` refused every spec outside
756 // `1..=MAX_FRAME_CHANNELS` before this resampler existed.
757 self.target.channels() as u8,
758 self.target_format,
759 self.target_layout.clone(),
760 planes,
761 plane_count as u8,
762 AudioFrameExtra::default(),
763 )
764 .with_pts(Some(Timestamp::new(pts, self.target_timebase)))
765 .with_duration(Some(Timestamp::new(produced as i64, self.target_timebase))),
766 )
767 }
768}
769
770/// Everything a converted frame needs, acquired before the conversion
771/// runs. See [`FfmpegResampler::prepare_output`].
772struct PreparedOutput<C: crate::FfmpegCarrier + crate::CarrierOps> {
773 /// The output `AVFrame`. Owning it here is what keeps the plane
774 /// pointers below valid across the conversion — moving this struct
775 /// moves a pointer to the `AVFrame`, never the `AVFrame` or the
776 /// buffers it addresses.
777 frame: frame::Audio,
778 /// One carrier claim per populated plane, taken before the
779 /// conversion ran and settled at its true length after — which is
780 /// what keeps this struct's whole reason for existing intact on the
781 /// view lane too. `None` for every unpopulated slot.
782 reserved: [Option<C::Reserved>; 8],
783 plane_count: usize,
784 /// Bytes one plane holds at full capacity — the ceiling every trim
785 /// stays under.
786 plane_len: usize,
787 /// Bytes one plane holds per sample.
788 per_sample: usize,
789}
790
791impl<C: crate::FfmpegCarrier + crate::CarrierOps> CarrierResampler<C> {
792 /// **Always [`Sent::Accepted`] when it accepts at all.** The
793 /// converted-frame queue this type keeps is unbounded — every frame a
794 /// conversion produces is built before `swr` is asked and pushed
795 /// straight onto it — so there is no state in which draining first
796 /// would let a submission through that is refused now. A bounded
797 /// implementation of the same face would answer
798 /// [`Sent::MustDrain`] here; this one has nothing to say it about.
799 ///
800 /// [`ResampleError::AfterEof`] stays an error rather than becoming
801 /// that arm, and the line is the same one the decoders draw: a
802 /// resampler that has been told the stream ended will refuse this
803 /// frame however much is drained first, so sending the caller into a
804 /// drain loop would be sending it nowhere. `flush` is the way back,
805 /// and the message says so.
806 pub(crate) fn send_frame_impl(&mut self, frame: &Frame<C>) -> Result<Sent, ResampleError> {
807 if self.eof {
808 return Err(ResampleError::AfterEof);
809 }
810 self.check_source(frame)?;
811 // A frame carrying no samples is a header and nothing else. There
812 // is nothing to convert and nothing to stage: `av_frame_get_buffer`
813 // refuses a zero-sample allocation, so staging one would hand `swr`
814 // an unbacked `AVFrame` for no gain.
815 if frame.nb_samples() == 0 {
816 return Ok(Sent::Accepted);
817 }
818
819 // Nothing below touches the session's state until the conversion
820 // has succeeded. A refused frame must leave the timeline exactly
821 // where it was, or the next good frame inherits the rejected one's
822 // timestamp.
823 let anchor = self.anchor_of(frame)?;
824 let input = self.stage_input(frame)?;
825 let capacity = self.output_capacity(frame.nb_samples() as i64)?;
826 self.check_timeline(anchor, capacity)?;
827 let mut prepared = self.prepare_output(capacity)?;
828 // The last fallible thing before the conversion: room for the frame
829 // it will produce. `push_back` on a full queue allocates, and an
830 // allocation failure there aborts the process rather than
831 // unwinding — so the growth happens here, where it can be an error.
832 self
833 .ready
834 .try_reserve(1)
835 .map_err(|_| ResampleError::QueueAlloc)?;
836
837 // The only mutation. Everything above could fail and cost nothing;
838 // nothing below can fail at all.
839 self
840 .ctx
841 .run(&input, &mut prepared.frame)
842 .map_err(|e| ResampleError::Resample(Error::Ffmpeg(e)))?;
843
844 // The frame is inside the filter now, so the timeline may be
845 // anchored on it. Anchored on *input* rather than on the first
846 // output, because a call that produces nothing but fills the delay
847 // line still fixes where the stream starts.
848 if self.next_pts.is_none() {
849 self.next_pts = anchor;
850 }
851 if let Some(converted) = self.finish_output(prepared) {
852 self.ready.push_back(converted);
853 }
854 Ok(Sent::Accepted)
855 }
856
857 /// **No parked-frame seat here, and none is needed.** The queue holds
858 /// frames that are already built: every fallible step of a
859 /// conversion — the carrier's claim included — happens in
860 /// [`Self::prepare_output`], before `swr` consumes anything, and
861 /// `finish_output` is infallible by construction. There is no
862 /// conversion left to fail after a frame has been taken out of the
863 /// queue, so nothing can be lost between the two. That is the
864 /// property the reserve-then-commit seam was built for, stated where
865 /// the sibling roads state their seats.
866 pub(crate) fn receive_frame_impl(
867 &mut self,
868 dst: &mut Frame<C>,
869 ) -> Result<Received, ResampleError> {
870 if let Some(frame) = self.ready.pop_front() {
871 *dst = frame;
872 return Ok(Received::Frame);
873 }
874 if !self.eof {
875 return Ok(Received::NeedsInput);
876 }
877 if self.drained {
878 return Ok(Received::Ended);
879 }
880 // EOF: drain the conversion tail. Without this every file loses the
881 // tens of milliseconds sitting inside the filter.
882 //
883 // **This is where the two answers used to be one.** Pre-EOF nothing
884 // ready and post-EOF tail exhausted both returned `Again`, so a
885 // caller that did not itself remember whether it had called
886 // `send_eof` could not tell "send more" from "there is no more" —
887 // and a drain loop written against the seam alone spun forever on
888 // a resampler that was already finished.
889 let remaining = self.delay_impl();
890 if remaining <= 0 {
891 self.drained = true;
892 return Ok(Received::Ended);
893 }
894 let capacity = remaining.min(i64::from(i32::MAX)) as usize;
895 // Same discipline as `send_frame`, and for the same reason: the
896 // tail is drained only once the timeline can hold it and every
897 // reference the converted frame needs is already in hand, so a
898 // failure leaves the delay line untouched instead of turning
899 // samples into an error.
900 self.check_timeline(None, capacity)?;
901 let mut prepared = self.prepare_output(capacity)?;
902 self
903 .ctx
904 .flush(&mut prepared.frame)
905 .map_err(|e| ResampleError::Resample(Error::Ffmpeg(e)))?;
906 match self.finish_output(prepared) {
907 Some(frame) => {
908 *dst = frame;
909 Ok(Received::Frame)
910 }
911 // The delay line reported samples and the flush produced none.
912 // Another flush would report the same and produce the same — this
913 // really happens, `swr_get_delay` standing at a residue the
914 // converter will not emit — so this is the end of the tail and
915 // not a pause in it. Saying otherwise is the spin this reform
916 // removes.
917 None => {
918 self.drained = true;
919 Ok(Received::Ended)
920 }
921 }
922 }
923
924 pub(crate) fn send_eof_impl(&mut self) -> Result<Sent, ResampleError> {
925 self.eof = true;
926 Ok(Sent::Accepted)
927 }
928
929 /// Resets the resampler for another stream on the same two specs.
930 ///
931 /// The `swr` context is **rebuilt**, not drained. `swresample` has no
932 /// reset call, and draining it dry cannot be verified from outside: a
933 /// `swr_convert_frame` that makes no progress reports no error, so a
934 /// drain loop that gives up and a drain loop that finished are
935 /// indistinguishable — and a flush that returned `Ok` with the old
936 /// delay line still inside would let one stream's tail contaminate
937 /// the next. A fresh context is the only reset whose success is a
938 /// fact.
939 ///
940 /// The new context is built before the old one is dropped, so a
941 /// failure leaves the resampler exactly as it was: this call either
942 /// resets everything or changes nothing.
943 pub(crate) fn flush_impl(&mut self) -> Result<(), ResampleError> {
944 let ctx = open_context(
945 &self.source,
946 &self.target,
947 self.staged_source_layout,
948 self.staged_target_layout,
949 )?;
950 self.ctx = ctx;
951 self.ready.clear();
952 self.next_pts = None;
953 self.eof = false;
954 self.drained = false;
955 debug_assert_eq!(self.delay_impl(), 0, "a fresh swr context holds nothing");
956 Ok(())
957 }
958}
959
960macro_rules! resampler_lane_face {
961 ($($lane:ty),+ $(,)?) => { $(
962 impl CarrierResampler<$lane> {
963 /// Opens a resampler between two explicit specs. See
964 /// [`CarrierResampler::new_impl`] for the full contract.
965 pub fn new(
966 source: ResampleSpec,
967 target: ResampleSpec,
968 limits: FrameLimits,
969 ) -> Result<Self, ResampleError> {
970 Self::new_impl(source, target, limits)
971 }
972
973 /// The spec frames must arrive in.
974 pub const fn source(&self) -> &ResampleSpec {
975 self.source_impl()
976 }
977
978 /// The spec frames leave in.
979 pub const fn target(&self) -> &ResampleSpec {
980 self.target_impl()
981 }
982
983 /// The wrapped `swr` context.
984 pub const fn inner(&self) -> &resampling::Context {
985 self.inner_impl()
986 }
987
988 /// Samples still inside the delay line, counted at the output
989 /// rate.
990 pub fn delay(&self) -> i64 {
991 self.delay_impl()
992 }
993 }
994
995 impl AudioResampler for CarrierResampler<$lane> {
996 type Adapter = Ffmpeg;
997 type Buffer = <$lane as crate::FfmpegCarrier>::Buffer;
998 type Error = ResampleError;
999
1000 fn send_frame(&mut self, frame: &Frame<$lane>) -> Result<Sent, ResampleError> {
1001 self.send_frame_impl(frame)
1002 }
1003
1004 fn receive_frame(&mut self, dst: &mut Frame<$lane>) -> Result<Received, ResampleError> {
1005 self.receive_frame_impl(dst)
1006 }
1007
1008 fn send_eof(&mut self) -> Result<Sent, ResampleError> {
1009 self.send_eof_impl()
1010 }
1011
1012 fn flush(&mut self) -> Result<(), ResampleError> {
1013 self.flush_impl()
1014 }
1015 }
1016 )+ };
1017}
1018
1019resampler_lane_face!(crate::View, crate::Owned);
1020
1021/// Payload for [`ResampleError::SourceChanged`].
1022///
1023/// A frame arrived whose shape is not the source spec this resampler
1024/// was built with — the mid-stream refusal.
1025///
1026/// The face never silently reconfigures: doing so would resample the
1027/// two halves of a stream on different terms and hand back a single
1028/// unbroken timeline built out of them. Build a new resampler for the
1029/// new source spec.
1030#[derive(thiserror::Error, Debug, Clone)]
1031#[error(
1032 "source format changed mid-stream: expected {expected_rate} Hz {expected_format:?}, \
1033 got {found_rate} Hz {found_format:?}"
1034)]
1035pub struct SourceChanged {
1036 expected_rate: u32,
1037 expected_format: SampleFormat,
1038 found_rate: u32,
1039 found_format: SampleFormat,
1040}
1041
1042impl SourceChanged {
1043 /// Constructs a `SourceChanged` payload.
1044 #[inline]
1045 pub const fn new(
1046 expected_rate: u32,
1047 expected_format: SampleFormat,
1048 found_rate: u32,
1049 found_format: SampleFormat,
1050 ) -> Self {
1051 Self {
1052 expected_rate,
1053 expected_format,
1054 found_rate,
1055 found_format,
1056 }
1057 }
1058 /// Rate the resampler was built for.
1059 #[inline]
1060 pub const fn expected_rate(&self) -> u32 {
1061 self.expected_rate
1062 }
1063 /// Sample format the resampler was built for.
1064 #[inline]
1065 pub const fn expected_format(&self) -> SampleFormat {
1066 self.expected_format
1067 }
1068 /// Rate the offending frame carried.
1069 #[inline]
1070 pub const fn found_rate(&self) -> u32 {
1071 self.found_rate
1072 }
1073 /// Sample format the offending frame carried.
1074 #[inline]
1075 pub const fn found_format(&self) -> SampleFormat {
1076 self.found_format
1077 }
1078}
1079
1080/// Payload for [`ResampleError::PlaneCount`].
1081///
1082/// A frame's planes do not hold what its header claims — too few
1083/// planes for the format, or a plane shorter than its sample count
1084/// requires.
1085#[derive(thiserror::Error, Debug, Clone)]
1086#[error("frame plane geometry mismatch: expected {expected}, found {found}")]
1087pub struct PlaneCount {
1088 expected: usize,
1089 found: usize,
1090}
1091
1092impl PlaneCount {
1093 /// Constructs a `PlaneCount` payload.
1094 #[inline]
1095 pub const fn new(expected: usize, found: usize) -> Self {
1096 Self { expected, found }
1097 }
1098 /// What the format and sample count require.
1099 #[inline]
1100 pub const fn expected(&self) -> usize {
1101 self.expected
1102 }
1103 /// What the frame carries.
1104 #[inline]
1105 pub const fn found(&self) -> usize {
1106 self.found
1107 }
1108}
1109
1110/// Payload for [`ResampleError::OutputTooLarge`].
1111///
1112/// The converted frame would be larger than
1113/// [`FrameLimits::max_frame_bytes`] allows.
1114///
1115/// Distinct from [`SampleCount`], which is about a count no `AVFrame`
1116/// can express at all. This one is about a frame FFmpeg would happily
1117/// allocate and the caller has said it does not want: the amplification
1118/// a hostile source rate buys — one second at a declared 1 Hz becomes
1119/// gigabytes at 192 kHz — lands here.
1120#[derive(thiserror::Error, Debug, Clone, Copy, PartialEq, Eq)]
1121#[error("a converted frame of {bytes} bytes exceeds the {limit}-byte ceiling")]
1122pub struct OutputTooLarge {
1123 bytes: usize,
1124 limit: usize,
1125}
1126
1127impl OutputTooLarge {
1128 /// Constructs an `OutputTooLarge` payload.
1129 #[cfg_attr(not(tarpaulin), inline(always))]
1130 pub const fn new(bytes: usize, limit: usize) -> Self {
1131 Self { bytes, limit }
1132 }
1133 /// The bytes the conversion would have produced.
1134 #[cfg_attr(not(tarpaulin), inline(always))]
1135 pub const fn bytes(&self) -> usize {
1136 self.bytes
1137 }
1138 /// The ceiling in force.
1139 #[cfg_attr(not(tarpaulin), inline(always))]
1140 pub const fn limit(&self) -> usize {
1141 self.limit
1142 }
1143}
1144
1145/// Payload for [`ResampleError::SampleCount`].
1146///
1147/// A sample count no frame can hold: one whose byte size overflows, or
1148/// one past the `c_int` `av_frame_get_buffer` takes.
1149#[derive(thiserror::Error, Debug, Clone)]
1150#[error("{requested} samples is not a frame size")]
1151pub struct SampleCount {
1152 requested: usize,
1153}
1154
1155impl SampleCount {
1156 /// Constructs a `SampleCount` payload.
1157 #[inline]
1158 pub const fn new(requested: usize) -> Self {
1159 Self { requested }
1160 }
1161 /// The count that was asked for.
1162 #[inline]
1163 pub const fn requested(&self) -> usize {
1164 self.requested
1165 }
1166}
1167
1168/// Payload for [`ResampleError::UnsupportedRate`].
1169///
1170/// One end of the conversion declares a sample rate `swr` cannot be
1171/// driven with — zero, or past `c_int`.
1172#[derive(thiserror::Error, Debug, Clone)]
1173#[error("the {end} rate {rate} is not a sample rate swr can use")]
1174pub struct UnsupportedRate {
1175 end: SpecEnd,
1176 rate: u32,
1177}
1178
1179impl UnsupportedRate {
1180 /// Constructs an `UnsupportedRate` payload.
1181 #[inline]
1182 pub const fn new(end: SpecEnd, rate: u32) -> Self {
1183 Self { end, rate }
1184 }
1185 /// Which end of the conversion.
1186 #[inline]
1187 pub const fn end(&self) -> SpecEnd {
1188 self.end
1189 }
1190 /// The rate that was declared.
1191 #[inline]
1192 pub const fn rate(&self) -> u32 {
1193 self.rate
1194 }
1195}
1196
1197/// Payload for [`ResampleError::UnsupportedFormat`].
1198///
1199/// One end of the conversion declares no sample format
1200/// (`AV_SAMPLE_FMT_NONE`) — the state a codec context is in before its
1201/// decoder opens.
1202#[derive(thiserror::Error, Debug, Clone)]
1203#[error("the {end} spec names no sample format")]
1204pub struct UnsupportedFormat {
1205 end: SpecEnd,
1206}
1207
1208impl UnsupportedFormat {
1209 /// Constructs an `UnsupportedFormat` payload.
1210 #[inline]
1211 pub const fn new(end: SpecEnd) -> Self {
1212 Self { end }
1213 }
1214 /// Which end of the conversion.
1215 #[inline]
1216 pub const fn end(&self) -> SpecEnd {
1217 self.end
1218 }
1219}
1220
1221/// Payload for [`ResampleError::UnsupportedLayout`].
1222///
1223/// One end of the conversion declares a channel layout this backend
1224/// will not carry.
1225///
1226/// Native and unspecified layouts are the two it does. A **custom** or
1227/// **ambisonic** `AVChannelLayout` owns a heap-allocated channel map,
1228/// and FFmpeg documents that such a layout must be copied with
1229/// `av_channel_layout_copy` rather than assigned — while
1230/// `ffmpeg_next::ChannelLayout` is a `Copy` wrapper with no destructor.
1231/// Every `AVFrame` this type stages or allocates receives the layout by
1232/// assignment, and `av_frame_free` runs `av_channel_layout_uninit` on
1233/// it: the first staged frame to be dropped would free a map the spec,
1234/// the decoder and every later frame still point at. Refusing at
1235/// construction is what keeps that use-after-free unreachable; a
1236/// resampler over those layouts is a separate design, not a silent
1237/// approximation.
1238#[derive(thiserror::Error, Debug, Clone)]
1239#[error("the {end} channel layout is not supported: order {order}, {channels} channels")]
1240pub struct UnsupportedLayout {
1241 end: SpecEnd,
1242 order: i32,
1243 channels: i32,
1244}
1245
1246impl UnsupportedLayout {
1247 /// Constructs an `UnsupportedLayout` payload.
1248 #[inline]
1249 pub const fn new(end: SpecEnd, order: i32, channels: i32) -> Self {
1250 Self {
1251 end,
1252 order,
1253 channels,
1254 }
1255 }
1256 /// Which end of the conversion.
1257 #[inline]
1258 pub const fn end(&self) -> SpecEnd {
1259 self.end
1260 }
1261 /// `AVChannelOrder` as the raw integer it is on the wire.
1262 #[inline]
1263 pub const fn order(&self) -> i32 {
1264 self.order
1265 }
1266 /// The channel count the layout declares.
1267 #[inline]
1268 pub const fn channels(&self) -> i32 {
1269 self.channels
1270 }
1271}
1272
1273/// Payload for [`ResampleError::TooManyPlanes`].
1274///
1275/// A planar spec with more channels than a decoded frame has plane
1276/// slots.
1277///
1278/// `mediadecode`'s `AudioFrame` carries a fixed eight planes
1279/// (`AV_NUM_DATA_POINTERS`); planar audio past that lives in
1280/// `AVFrame.extended_data[]`, which this crate does not plumb through.
1281/// As a **source** no valid frame could ever arrive; as a **target**
1282/// `swr` would produce one this crate cannot hand back — and it would
1283/// fail only after the input had been consumed, leaving a session that
1284/// cannot be retried. Both are refused at construction, where nothing
1285/// has happened yet.
1286#[derive(thiserror::Error, Debug, Clone)]
1287#[error("the {end} spec is planar with {channels} channels; a frame carries {limit} planes")]
1288pub struct TooManyPlanes {
1289 end: SpecEnd,
1290 channels: i32,
1291 limit: i32,
1292}
1293
1294impl TooManyPlanes {
1295 /// Constructs a `TooManyPlanes` payload.
1296 #[inline]
1297 pub const fn new(end: SpecEnd, channels: i32, limit: i32) -> Self {
1298 Self {
1299 end,
1300 channels,
1301 limit,
1302 }
1303 }
1304 /// Which end of the conversion.
1305 #[inline]
1306 pub const fn end(&self) -> SpecEnd {
1307 self.end
1308 }
1309 /// The channel count the layout declares.
1310 #[inline]
1311 pub const fn channels(&self) -> i32 {
1312 self.channels
1313 }
1314 /// Plane slots a frame has.
1315 #[inline]
1316 pub const fn limit(&self) -> i32 {
1317 self.limit
1318 }
1319}
1320
1321/// Payload for [`ResampleError::UnsupportedChannelCount`].
1322///
1323/// A spec declaring more channels than a frame's channel seat can
1324/// carry. `mediadecode`'s `AudioFrame` states its channel count in a
1325/// `u8`, so 255 is the ceiling in both directions.
1326///
1327/// This is the **packed** sibling of [`TooManyPlanes`], which only ever
1328/// caught planar specs: packed audio declares one plane whatever its
1329/// channel count, so a 256-channel packed spec sailed past that check
1330/// and had its count clipped on the way into the frame — a frame whose
1331/// bytes were computed from 256 channels while advertising 255. Refused
1332/// here, at the same choke point, for both ends.
1333#[derive(thiserror::Error, Debug, Clone)]
1334#[error("the {end} spec declares {channels} channels; a frame carries at most {limit}")]
1335pub struct UnsupportedChannelCount {
1336 end: SpecEnd,
1337 channels: i32,
1338 limit: i32,
1339}
1340
1341impl UnsupportedChannelCount {
1342 /// Constructs an `UnsupportedChannelCount` payload.
1343 #[inline]
1344 pub const fn new(end: SpecEnd, channels: i32, limit: i32) -> Self {
1345 Self {
1346 end,
1347 channels,
1348 limit,
1349 }
1350 }
1351 /// Which end of the conversion.
1352 #[inline]
1353 pub const fn end(&self) -> SpecEnd {
1354 self.end
1355 }
1356 /// The channel count the layout declares.
1357 #[inline]
1358 pub const fn channels(&self) -> i32 {
1359 self.channels
1360 }
1361 /// Channels a frame can state.
1362 #[inline]
1363 pub const fn limit(&self) -> i32 {
1364 self.limit
1365 }
1366}
1367
1368/// Payload for [`ResampleError::TimestampOutOfRange`].
1369///
1370/// A frame's timestamp does not land on the output timeline: it does
1371/// not survive the rescale as an `i64`, or it is `AV_NOPTS_VALUE`,
1372/// which is a sentinel rather than a time.
1373///
1374/// Raised before anything is staged, so a refused frame leaves the
1375/// resampler exactly as it was.
1376#[derive(thiserror::Error, Debug, Clone)]
1377#[error("the frame timestamp {pts} does not land on the output timeline")]
1378pub struct TimestampOutOfRange {
1379 pts: i64,
1380}
1381
1382impl TimestampOutOfRange {
1383 /// Constructs a `TimestampOutOfRange` payload.
1384 #[inline]
1385 pub const fn new(pts: i64) -> Self {
1386 Self { pts }
1387 }
1388 /// The timestamp the frame carried, in its own timebase.
1389 #[inline]
1390 pub const fn pts(&self) -> i64 {
1391 self.pts
1392 }
1393}
1394
1395/// Payload for [`ResampleError::ChannelDropped`].
1396///
1397/// The conversion between these two layouts would silently drop a
1398/// source channel: FFmpeg's own mixing matrix routes it to no output.
1399///
1400/// `swr` mixes the channel positions its rematrix table knows and
1401/// processes the rest of the input as though it were absent — a log
1402/// line at most. Measured against FFmpeg 9, packed 22.2 → mono loses
1403/// fifteen of twenty-four channels and `cube` → stereo loses two of
1404/// eight, so this is not a matter of channel count. Installing an
1405/// explicit mix matrix is how such a conversion would be accepted
1406/// deliberately; until this crate has a seat for one, the pair is
1407/// refused.
1408#[derive(thiserror::Error, Debug, Clone)]
1409#[error(
1410 "converting {source_channels} channels to {target_channels} would drop source channel \
1411 {channel}: FFmpeg's mixing matrix routes it to no output"
1412)]
1413pub struct ChannelDropped {
1414 source_channels: i32,
1415 target_channels: i32,
1416 channel: i32,
1417}
1418
1419impl ChannelDropped {
1420 /// Constructs a `ChannelDropped` payload.
1421 #[inline]
1422 pub const fn new(source_channels: i32, target_channels: i32, channel: i32) -> Self {
1423 Self {
1424 source_channels,
1425 target_channels,
1426 channel,
1427 }
1428 }
1429 /// Channels the source layout declares.
1430 #[inline]
1431 pub const fn source_channels(&self) -> i32 {
1432 self.source_channels
1433 }
1434 /// Channels the target layout declares.
1435 #[inline]
1436 pub const fn target_channels(&self) -> i32 {
1437 self.target_channels
1438 }
1439 /// The first source channel that reaches no output channel.
1440 #[inline]
1441 pub const fn channel(&self) -> i32 {
1442 self.channel
1443 }
1444}
1445
1446/// Payload for [`ResampleError::RematrixUnsupported`].
1447///
1448/// FFmpeg will not build a mixing matrix between these two layouts at
1449/// all.
1450#[derive(thiserror::Error, Debug, Clone)]
1451#[error("FFmpeg builds no mixing matrix from {source_channels} channels to {target_channels}")]
1452pub struct RematrixUnsupported {
1453 source_channels: i32,
1454 target_channels: i32,
1455}
1456
1457impl RematrixUnsupported {
1458 /// Constructs a `RematrixUnsupported` payload.
1459 #[inline]
1460 pub const fn new(source_channels: i32, target_channels: i32) -> Self {
1461 Self {
1462 source_channels,
1463 target_channels,
1464 }
1465 }
1466 /// Channels the source layout declares.
1467 #[inline]
1468 pub const fn source_channels(&self) -> i32 {
1469 self.source_channels
1470 }
1471 /// Channels the target layout declares.
1472 #[inline]
1473 pub const fn target_channels(&self) -> i32 {
1474 self.target_channels
1475 }
1476}
1477
1478/// Payload for [`ResampleError::TimestampOverflow`].
1479///
1480/// The output timeline would leave `i64`. Counted timestamps are exact
1481/// or they are nothing, so this is named rather than saturated.
1482#[derive(thiserror::Error, Debug, Clone)]
1483#[error("the output timeline overflows: {pts} + {samples} samples")]
1484pub struct TimestampOverflow {
1485 pts: i64,
1486 samples: i64,
1487}
1488
1489impl TimestampOverflow {
1490 /// Constructs a `TimestampOverflow` payload.
1491 #[inline]
1492 pub const fn new(pts: i64, samples: i64) -> Self {
1493 Self { pts, samples }
1494 }
1495 /// Where the timeline stood.
1496 #[inline]
1497 pub const fn pts(&self) -> i64 {
1498 self.pts
1499 }
1500 /// How many samples were produced.
1501 #[inline]
1502 pub const fn samples(&self) -> i64 {
1503 self.samples
1504 }
1505}
1506
1507/// Payload for [`ResampleError::OutputBuffer`].
1508///
1509/// A reference to one of the output frame's planes could not be taken.
1510///
1511/// Raised while preparing the conversion, never after it: that is the
1512/// point of preparing.
1513#[derive(thiserror::Error, Debug, Clone)]
1514#[error("the output frame's plane {plane} could not be referenced")]
1515pub struct OutputBuffer {
1516 plane: usize,
1517}
1518
1519impl OutputBuffer {
1520 /// Constructs an `OutputBuffer` payload.
1521 #[inline]
1522 pub const fn new(plane: usize) -> Self {
1523 Self { plane }
1524 }
1525 /// Which plane slot.
1526 #[inline]
1527 pub const fn plane(&self) -> usize {
1528 self.plane
1529 }
1530}
1531
1532/// Errors from [`FfmpegResampler`] — **faults and the two send-side
1533/// refusals** ([`Self::SourceChanged`], [`Self::AfterEof`]).
1534///
1535/// `Again` used to be here and meant two opposite things: pre-EOF
1536/// "nothing ready, send more" and post-EOF "the tail is exhausted".
1537/// A caller polling the seam could tell them apart only by remembering
1538/// whether it had itself called `send_eof`; one that did not spun
1539/// forever. Both are [`Received`] states now, and they are distinct.
1540///
1541/// **Open fault taxonomy, so it is `#[non_exhaustive]`.** New ways to
1542/// fail are discovered — a backend, a ceiling, a corruption a codec
1543/// learns to report — and a consumer that meets one it has never heard
1544/// of should take its generic-fault path. That is exactly what the
1545/// wildcard arm this attribute forces is for. The two status
1546/// vocabularies opposite it,
1547/// [`Sent`](mediadecode::Sent) and [`Received`](mediadecode::Received),
1548/// are exhaustive for the mirror-image reason: their arms are the
1549/// substrate's fixed state set, and there the wildcard would be dead
1550/// weight hiding a state a consumer forgot.
1551#[derive(thiserror::Error, Debug, Clone, IsVariant, Unwrap, TryUnwrap)]
1552#[unwrap(ref, ref_mut)]
1553#[try_unwrap(ref, ref_mut)]
1554#[non_exhaustive]
1555pub enum ResampleError {
1556 /// The conversion would produce a frame larger than the ceiling
1557 /// allows. Refused **before** the output frame is allocated.
1558 #[error(transparent)]
1559 OutputTooLarge(#[from] OutputTooLarge),
1560
1561 /// A frame arrived whose shape is not the source spec this resampler
1562 /// was built with — the mid-stream refusal.
1563 #[error(transparent)]
1564 SourceChanged(#[from] SourceChanged),
1565
1566 /// [`send_frame`](AudioResampler::send_frame) was called after
1567 /// [`send_eof`](AudioResampler::send_eof). Call
1568 /// [`flush`](AudioResampler::flush) first to reuse the resampler for
1569 /// another stream.
1570 #[error("send_frame after send_eof; flush() first to start another stream")]
1571 AfterEof,
1572
1573 /// A frame's planes do not hold what its header claims — too few
1574 /// planes for the format, or a plane shorter than its sample count
1575 /// requires.
1576 #[error(transparent)]
1577 PlaneCount(#[from] PlaneCount),
1578
1579 /// A sample count no frame can hold: one whose byte size overflows,
1580 /// or one past the `c_int` `av_frame_get_buffer` takes.
1581 #[error(transparent)]
1582 SampleCount(#[from] SampleCount),
1583
1584 /// One end of the conversion declares a sample rate `swr` cannot be
1585 /// driven with — zero, or past `c_int`.
1586 #[error(transparent)]
1587 UnsupportedRate(#[from] UnsupportedRate),
1588
1589 /// One end of the conversion declares no sample format
1590 /// (`AV_SAMPLE_FMT_NONE`) — the state a codec context is in before
1591 /// its decoder opens.
1592 #[error(transparent)]
1593 UnsupportedFormat(#[from] UnsupportedFormat),
1594
1595 /// One end of the conversion declares a channel layout this backend
1596 /// will not carry.
1597 #[error(transparent)]
1598 UnsupportedLayout(#[from] UnsupportedLayout),
1599
1600 /// A planar spec with more channels than a decoded frame has plane
1601 /// slots.
1602 #[error(transparent)]
1603 TooManyPlanes(#[from] TooManyPlanes),
1604
1605 /// A spec with more channels than a frame's channel seat can state.
1606 #[error(transparent)]
1607 UnsupportedChannelCount(#[from] UnsupportedChannelCount),
1608
1609 /// A frame's timestamp does not land on the output timeline.
1610 #[error(transparent)]
1611 TimestampOutOfRange(#[from] TimestampOutOfRange),
1612
1613 /// The conversion between these two layouts would silently drop a
1614 /// source channel.
1615 #[error(transparent)]
1616 ChannelDropped(#[from] ChannelDropped),
1617
1618 /// FFmpeg will not build a mixing matrix between these two layouts at
1619 /// all.
1620 #[error(transparent)]
1621 RematrixUnsupported(#[from] RematrixUnsupported),
1622
1623 /// The output timeline would leave `i64`. Counted timestamps are
1624 /// exact or they are nothing, so this is named rather than saturated.
1625 #[error(transparent)]
1626 TimestampOverflow(#[from] TimestampOverflow),
1627
1628 /// The wrapped `swresample` call reported an error.
1629 #[error(transparent)]
1630 Resample(#[from] Error),
1631
1632 /// A reference to one of the output frame's planes could not be
1633 /// taken.
1634 #[error(transparent)]
1635 OutputBuffer(#[from] OutputBuffer),
1636
1637 /// The queue of converted frames could not be grown to hold one more.
1638 #[error("out of memory reserving room for a converted frame")]
1639 QueueAlloc,
1640}
1641
1642/// Which end of a conversion a refusal is about.
1643#[derive(Copy, Clone, Debug, PartialEq, Eq, IsVariant)]
1644pub enum SpecEnd {
1645 /// The spec frames must arrive in.
1646 Source,
1647 /// The spec frames leave in.
1648 Target,
1649}
1650
1651impl core::fmt::Display for SpecEnd {
1652 fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
1653 f.write_str(match self {
1654 Self::Source => "source",
1655 Self::Target => "target",
1656 })
1657 }
1658}
1659
1660/// Plane slots a `mediadecode::frame::AudioFrame` has — the fixed array
1661/// matching `AV_NUM_DATA_POINTERS`. Planar audio past this many
1662/// channels lives in `AVFrame.extended_data[]` / `extended_buf[]`,
1663/// which this crate does not plumb through: `convert` refuses such a
1664/// frame and `AudioFrame::new` will not build one.
1665const MAX_AUDIO_PLANES: i32 = 8;
1666
1667/// Channels a `mediadecode::frame::AudioFrame` can state — its channel
1668/// seat is a `u8`. A spec past this is refused rather than clipped.
1669const MAX_FRAME_CHANNELS: i32 = u8::MAX as i32;
1670
1671/// Refuses a spec `swr` cannot be driven with, or whose channel layout
1672/// cannot be carried by value — see [`ResampleError::UnsupportedLayout`]
1673/// for that one, which is the whole reason this check exists at the
1674/// choke point rather than in the `const` constructor.
1675fn check_spec(spec: &ResampleSpec, end: SpecEnd) -> Result<(), ResampleError> {
1676 if spec.rate == 0 || spec.rate > i32::MAX as u32 {
1677 return Err(ResampleError::UnsupportedRate(UnsupportedRate::new(
1678 end, spec.rate,
1679 )));
1680 }
1681 if spec.format == Sample::None {
1682 return Err(ResampleError::UnsupportedFormat(UnsupportedFormat::new(
1683 end,
1684 )));
1685 }
1686 let order = layout_order(&spec.layout);
1687 let channels = spec.layout.channels();
1688 let carried = order == AVChannelOrder::AV_CHANNEL_ORDER_NATIVE as i32
1689 || order == AVChannelOrder::AV_CHANNEL_ORDER_UNSPEC as i32;
1690 if !carried || channels <= 0 {
1691 return Err(ResampleError::UnsupportedLayout(UnsupportedLayout::new(
1692 end, order, channels,
1693 )));
1694 }
1695 // A planar spec with more channels than the frame model has plane
1696 // slots is a resampler that cannot work in either direction, and
1697 // saying so here is the difference between a refusal at construction
1698 // and a refusal on every frame — the target one arriving *after*
1699 // `swr` has already consumed the input, which is not a state a caller
1700 // can retry from.
1701 if spec.format.is_planar() && channels > MAX_AUDIO_PLANES {
1702 return Err(ResampleError::TooManyPlanes(TooManyPlanes::new(
1703 end,
1704 channels,
1705 MAX_AUDIO_PLANES,
1706 )));
1707 }
1708 // And the packed sibling, which the plane check above cannot see: a
1709 // packed spec declares one plane at any channel count, so it reached
1710 // the frame with its count clipped to 255 instead of refused. That is
1711 // the same silent truncation the decode path was carrying, on the
1712 // other audio road — closed here, at the same choke point, so that
1713 // every channel count downstream is exact by construction.
1714 if channels > MAX_FRAME_CHANNELS {
1715 return Err(ResampleError::UnsupportedChannelCount(
1716 UnsupportedChannelCount::new(end, channels, MAX_FRAME_CHANNELS),
1717 ));
1718 }
1719 Ok(())
1720}
1721
1722/// A layout's `AVChannelOrder` as the integer it is on the wire.
1723///
1724/// Read raw rather than matched as an `AVChannelOrder`, the discipline
1725/// this crate keeps everywhere it touches a bindgen enum: a value
1726/// outside this build's discriminant set would be undefined behaviour
1727/// the moment it existed as one.
1728fn layout_order(layout: &ChannelLayout) -> i32 {
1729 // SAFETY: `layout` is a live `ChannelLayout` for the duration of this
1730 // call; `addr_of!` reaches its `order` field without forming a
1731 // reference to the enum.
1732 unsafe { read_unaligned(addr_of!(layout.0.order).cast::<i32>()) }
1733}
1734
1735/// FFmpeg's `SWR_CH_MAX`: the square its own matrix builder writes,
1736/// whatever the two layouts' channel counts are.
1737///
1738/// Not a convenience. `swr_build_matrix2` copies its internal
1739/// `[SWR_CH_MAX][SWR_CH_MAX]` block out at the caller's stride, so a
1740/// buffer sized to the actual channel counts is written far past its
1741/// end — measured, and the measurement is a killed process.
1742const SWR_CH_MAX: usize = 64;
1743
1744/// Refuses an **effective pair** whose rematrixing would silently drop
1745/// input channels.
1746///
1747/// Takes the layouts `swr` is configured with, not the ones the caller
1748/// declared. The two differ exactly where it matters: an unspecified
1749/// layout is resolved to FFmpeg's default for its channel count before
1750/// the context is opened, and twenty-four unspecified channels resolve
1751/// to 22.2 — so the declared pair says "unspecified, nothing to
1752/// rematrix" while the conversion that runs is the lossy one.
1753///
1754/// This is the second half of the crate's two-layout bookkeeping, and
1755/// the halves answer different questions. The **declared** layout is
1756/// what decoded frames carry (a WAV without a channel mask hands out
1757/// unspecified frames forever) and stays the yardstick for the
1758/// mid-stream refusal: *is this frame the stream I was built for?* The
1759/// **effective** layout is what `swr` and every staged `AVFrame` use,
1760/// and it is the one judged here: *what will `swr` actually do?*
1761///
1762/// Each end can be perfectly valid on its own and the conversion
1763/// between them still lose whole channels: `swr` mixes only the channel
1764/// positions its rematrix table knows, and quietly processes the rest
1765/// of the input as though it were not there. Measured against the
1766/// linked FFmpeg 9 with a tone isolated in each source channel: packed
1767/// 22.2 → mono drops fifteen of twenty-four (`swr` says as much in a
1768/// log line and converts anyway), `cube` → stereo drops two of *eight*
1769/// — so a channel-count threshold is both too strict and too loose to
1770/// be the rule.
1771///
1772/// The rule is asked of FFmpeg instead: build the mixing matrix its own
1773/// builder would use, and refuse when any input channel reaches no
1774/// output at all. `lfe_mix_level` is deliberately non-zero, so the
1775/// question is "can this channel reach the output" rather than "does
1776/// FFmpeg's default downmix policy include it" — the default leaves LFE
1777/// out of a downmix on purpose, and refusing an everyday 5.1 → stereo
1778/// over that would be absurd. The predicate matched the tone sweep
1779/// exactly on every pair measured.
1780///
1781/// A pair `swr` cannot matrix at all is refused too. Accepting these
1782/// deliberately is a *mix matrix* seat on the spec — a real design, not
1783/// something to mint in passing; until it exists, refusal is the honest
1784/// answer.
1785fn check_pair(source: &ChannelLayout, target: &ChannelLayout) -> Result<(), ResampleError> {
1786 let native = AVChannelOrder::AV_CHANNEL_ORDER_NATIVE as i32;
1787 // A layout still unspecified *after* resolution — a channel count
1788 // FFmpeg has no default for — is mapped positionally by `swr` with no
1789 // rematrixing at all, and identical layouts need no matrix: neither
1790 // can drop a channel, and neither is what the builder describes.
1791 if layout_order(source) != native || layout_order(target) != native || source == target {
1792 return Ok(());
1793 }
1794 let source_channels = source.channels();
1795 let target_channels = target.channels();
1796
1797 let mut matrix = vec![0f64; SWR_CH_MAX * SWR_CH_MAX];
1798 // SAFETY: both layouts are live for the call; `matrix` is the full
1799 // `SWR_CH_MAX` square the builder writes, passed with the matching
1800 // stride; the encoding is a compile-time constant of this build; and
1801 // a null log context is documented as allowed.
1802 let rc = unsafe {
1803 swr_build_matrix2(
1804 &source.0,
1805 &target.0,
1806 core::f64::consts::FRAC_1_SQRT_2,
1807 core::f64::consts::FRAC_1_SQRT_2,
1808 1.0,
1809 1.0,
1810 1.0,
1811 matrix.as_mut_ptr(),
1812 SWR_CH_MAX as isize,
1813 AVMatrixEncoding::AV_MATRIX_ENCODING_NONE,
1814 core::ptr::null_mut(),
1815 )
1816 };
1817 if rc < 0 {
1818 return Err(ResampleError::RematrixUnsupported(
1819 RematrixUnsupported::new(source_channels, target_channels),
1820 ));
1821 }
1822 for channel in 0..source_channels.min(SWR_CH_MAX as i32) {
1823 let index = channel as usize;
1824 if (0..target_channels.min(SWR_CH_MAX as i32) as usize)
1825 .all(|out| matrix[index + SWR_CH_MAX * out] == 0.0)
1826 {
1827 return Err(ResampleError::ChannelDropped(ChannelDropped::new(
1828 source_channels,
1829 target_channels,
1830 channel,
1831 )));
1832 }
1833 }
1834 Ok(())
1835}
1836
1837/// Opens a `swr` context for the two specs. Shared by
1838/// [`FfmpegResampler::new`] and the rebuild
1839/// [`AudioResampler::flush`] performs.
1840fn open_context(
1841 source: &ResampleSpec,
1842 target: &ResampleSpec,
1843 staged_source_layout: ChannelLayout,
1844 staged_target_layout: ChannelLayout,
1845) -> Result<resampling::Context, ResampleError> {
1846 resampling::Context::get(
1847 source.format,
1848 staged_source_layout,
1849 source.rate,
1850 target.format,
1851 staged_target_layout,
1852 target.rate,
1853 )
1854 .map_err(|e| ResampleError::Resample(Error::Ffmpeg(e)))
1855}
1856
1857/// Allocates an audio `AVFrame`, checking every step the dependency's
1858/// own `frame::Audio::new` does not.
1859///
1860/// `ffmpeg_next`'s constructor dereferences `av_frame_alloc`'s result
1861/// without a null check and discards `av_frame_get_buffer`'s return
1862/// value, so an allocation failure there yields a frame whose planes
1863/// are not backed — which is then handed to FFI. Both are checked here;
1864/// on failure the caller gets a named error and no frame at all. The
1865/// null check is the crate's existing one
1866/// ([`crate::frame::alloc_av_audio_frame`], which the decoders already
1867/// allocate through), so there is one answer to `av_frame_alloc`
1868/// returning null rather than two.
1869fn new_audio_frame(
1870 format: Sample,
1871 samples: usize,
1872 rate: u32,
1873 layout: ChannelLayout,
1874) -> Result<frame::Audio, ResampleError> {
1875 if samples == 0 || samples > i32::MAX as usize {
1876 return Err(ResampleError::SampleCount(SampleCount::new(samples)));
1877 }
1878 let mut out = crate::frame::alloc_av_audio_frame()?;
1879 out.set_format(format);
1880 out.set_samples(samples);
1881 // The layout is assigned by value, which is sound only because
1882 // `check_spec` refused every layout that owns a heap channel map.
1883 out.set_channel_layout(layout);
1884 out.set_rate(rate);
1885 // SAFETY: `out` is a live `AVFrame` whose format, sample count and
1886 // layout were just set; `av_frame_get_buffer` allocates its planes
1887 // and reports failure in its return value, which is checked.
1888 let rc = unsafe { av_frame_get_buffer(out.as_mut_ptr(), 0) };
1889 if rc < 0 {
1890 return Err(ResampleError::Resample(Error::Ffmpeg(
1891 ffmpeg_next::Error::from(rc),
1892 )));
1893 }
1894 Ok(out)
1895}
1896
1897/// Bytes one plane holds for `samples` samples of `format`, or `None`
1898/// when that product does not fit a `usize`. Packed formats keep every
1899/// channel in the single plane; planar formats give each channel its
1900/// own.
1901fn plane_bytes(format: Sample, samples: usize, channels: i32) -> Option<usize> {
1902 // Refused, not floored. `check_spec` already proves the count is in
1903 // `1..=MAX_FRAME_CHANNELS` before any resampler exists, so this is a
1904 // restatement of an invariant rather than a live branch — but it is
1905 // stated as a refusal because the alternative was a substituted `1`,
1906 // which invents a channel the caller never declared and makes the
1907 // byte product disagree with the frame it sizes.
1908 let channels = usize::try_from(channels).ok().filter(|count| *count > 0)?;
1909 let bytes = samples.checked_mul(format.bytes())?;
1910 if format.is_planar() {
1911 Some(bytes)
1912 } else {
1913 bytes.checked_mul(channels)
1914 }
1915}
1916
1917/// Builds a native-order [`ChannelLayout`] from a channel bitmask,
1918/// without ever forming an `AVChannelLayout` out of foreign memory:
1919/// the struct starts zeroed (`AV_CHANNEL_ORDER_UNSPEC` is `0`, a valid
1920/// discriminant) and FFmpeg fills it.
1921fn layout_from_mask(mask: u64) -> ChannelLayout {
1922 // SAFETY: a zeroed `AVChannelLayout` is a valid value — its `order`
1923 // field reads as `AV_CHANNEL_ORDER_UNSPEC`, the zero discriminant —
1924 // and `av_channel_layout_from_mask` overwrites it wholesale.
1925 unsafe {
1926 let mut layout = std::mem::zeroed();
1927 if av_channel_layout_from_mask(&mut layout, mask) < 0 {
1928 return ChannelLayout::default(mask.count_ones() as i32);
1929 }
1930 ChannelLayout(layout)
1931 }
1932}
1933
1934/// The layout `swr` will actually be configured with.
1935///
1936/// `swr_init` replaces an unspecified input or output layout with
1937/// FFmpeg's default for that channel count, and from then on compares
1938/// every frame handed to it against *that* layout — a staged frame
1939/// still carrying the unspecified one is refused with
1940/// `AVERROR_INPUT_CHANGED`. Applying the same rule here, once, keeps
1941/// the frames this type builds in step with the context it built.
1942///
1943/// The declared layout is kept separately and is what the mid-stream
1944/// check compares against, because it is what decoded frames really
1945/// carry: a WAV without a channel mask hands out unspecified frames
1946/// forever, whatever `swr` decided internally.
1947fn initialized_layout(layout: ChannelLayout) -> ChannelLayout {
1948 if layout.is_empty() {
1949 ChannelLayout::default(layout.channels())
1950 } else {
1951 layout
1952 }
1953}
1954
1955/// Reads an `AVChannelLayout` out of FFmpeg memory into a layout this
1956/// spec can own, or `None` for one it does not represent.
1957///
1958/// The `order` field is read as the integer it is on the wire: an
1959/// out-of-range value would be undefined behaviour the instant it
1960/// existed as an `AVChannelOrder`, which is the hazard this crate
1961/// keeps out everywhere it touches a bindgen enum.
1962///
1963/// A **custom** or **ambisonic** layout returns `None`. Both keep a
1964/// heap-allocated channel map inside the layout, and `ChannelLayout` is
1965/// a plain `Copy` wrapper with no destructor: owning one here would
1966/// either alias a map the decoder still frees or leak the copy. A
1967/// resampler over one of those layouts is a separate design, not a
1968/// silent approximation.
1969///
1970/// # Safety
1971///
1972/// `ptr` must be a live `*const AVChannelLayout` for the duration of
1973/// this call.
1974unsafe fn layout_from_raw(ptr: *const ffmpeg_next::ffi::AVChannelLayout) -> Option<ChannelLayout> {
1975 let order = unsafe { read_unaligned(addr_of!((*ptr).order).cast::<i32>()) };
1976 let channels = unsafe { (*ptr).nb_channels };
1977 if channels <= 0 {
1978 return None;
1979 }
1980 if order == AVChannelOrder::AV_CHANNEL_ORDER_NATIVE as i32 {
1981 // SAFETY: `u.mask` is the union's variant for NATIVE, and the
1982 // order was checked against our own constant before the read.
1983 let mask = unsafe { (*ptr).u.mask };
1984 if mask != 0 {
1985 return Some(layout_from_mask(mask));
1986 }
1987 // Native in name with no channels named: unspecified in substance.
1988 return Some(ResampleSpec::unspecified_layout(channels));
1989 }
1990 if order == AVChannelOrder::AV_CHANNEL_ORDER_UNSPEC as i32 {
1991 return Some(ResampleSpec::unspecified_layout(channels));
1992 }
1993 None
1994}
1995
1996/// Compile-time assurance that `SampleFormat`'s round trip through
1997/// FFmpeg's vocabulary is the identity on the closed set. Both
1998/// directions are hand-written tables, and a table that disagreed with
1999/// its inverse would silently mislabel every sample.
2000const _: () = {
2001 assert!(
2002 SampleFormat::from_raw(AVSampleFormat::AV_SAMPLE_FMT_NONE as i32)
2003 .to_ffmpeg()
2004 .is_none()
2005 );
2006};
2007
2008#[cfg(test)]
2009mod tests {
2010 use super::*;
2011
2012 use mediadecode::resampler::AudioResampler;
2013
2014 // These exercise the conversion arithmetic — rates, layouts, the
2015 // counted timeline, the byte ceiling — which is lane-independent, so
2016 // they run on the owned lane exactly as they did before the second
2017 // lane existed. The view lane's own road through this type (reserve
2018 // before `swr`, commit after) is proved in `tests/view_carriers.rs`,
2019 // where a produced plane can be shown to point into the output
2020 // frame's buffer.
2021 use crate::{FfmpegBytes, FfmpegOwnedResampler as FfmpegResampler};
2022
2023 type Frame = super::Frame<crate::Owned>;
2024
2025 /// A 48 kHz packed-s16 stereo frame of silence, with the plane its
2026 /// header claims.
2027 fn stereo_frame(samples: u32) -> Frame {
2028 let plane = FfmpegBytes::copy_from_slice(&vec![0u8; samples as usize * 2 * 2]);
2029 let planes = std::array::from_fn(|index| {
2030 Plane::new(
2031 if index == 0 {
2032 plane.clone()
2033 } else {
2034 FfmpegBytes::empty()
2035 },
2036 0,
2037 )
2038 });
2039 AudioFrame::new(
2040 48_000,
2041 samples,
2042 2,
2043 SampleFormat::S16,
2044 crate::channel_layout::channel_layout_description_from_ffmpeg(&ChannelLayout::STEREO),
2045 planes,
2046 1,
2047 AudioFrameExtra::default(),
2048 )
2049 .with_pts(Some(Timestamp::new(
2050 0,
2051 Timebase::new(1, std::num::NonZeroI32::new(48_000).expect("a real rate")),
2052 )))
2053 }
2054
2055 /// Takes every frame that is ready *right now* and stops — the
2056 /// pre-EOF half of a drain, written as the exhaustive match the face
2057 /// now requires. A `.is_ok()` loop here would never end: "needs
2058 /// input" is a success.
2059 fn drain_ready(resampler: &mut FfmpegResampler, dst: &mut Frame) {
2060 loop {
2061 match resampler.receive_frame(dst).expect("a fault-free drain") {
2062 Received::Frame => {}
2063 Received::NeedsInput | Received::Ended => return,
2064 }
2065 }
2066 }
2067
2068 fn stereo_to_mono() -> FfmpegResampler {
2069 FfmpegResampler::new(
2070 ResampleSpec::new(
2071 48_000,
2072 Sample::I16(ffmpeg_next::format::sample::Type::Packed),
2073 ChannelLayout::STEREO,
2074 ),
2075 ResampleSpec::new(
2076 16_000,
2077 Sample::I16(ffmpeg_next::format::sample::Type::Packed),
2078 ChannelLayout::MONO,
2079 ),
2080 FrameLimits::default(),
2081 )
2082 .expect("open resampler")
2083 }
2084
2085 #[test]
2086 fn resample_error_carries_the_derived_accessor_face() {
2087 // `IsVariant` / `Unwrap` / `TryUnwrap` — one arm per derive family,
2088 // mirroring the mediadecode-side proof for this crate's own
2089 // newly-wired `derive_more` dependency.
2090 let err = ResampleError::OutputBuffer(OutputBuffer::new(2));
2091 assert!(err.is_output_buffer());
2092 assert!(!err.is_queue_alloc());
2093 assert_eq!(err.unwrap_output_buffer_ref().plane(), 2);
2094 assert!(err.try_unwrap_queue_alloc().is_err());
2095 }
2096
2097 #[test]
2098 fn an_allocation_fault_while_sending_leaves_the_session_untouched() {
2099 // The class this design exists to end: a failure on the far side of
2100 // `swr_convert_frame` leaves a session no caller can act on —
2101 // retrying feeds the same samples twice, continuing loses them, and
2102 // the delay line has moved either way. Every allocation the
2103 // conversion needs is taken before `swr` runs, so an allocator that
2104 // refuses everything can only produce an error that cost nothing.
2105 crate::fault_subprocess::in_subprocess(
2106 "resampler::tests::an_allocation_fault_while_sending_leaves_the_session_untouched",
2107 || {
2108 let mut resampler = stereo_to_mono();
2109 let frame = stereo_frame(4_800);
2110 let mut dst = crate::boundary::empty_owned_audio_frame();
2111 crate::accepted(resampler.send_frame(&frame), "a first frame");
2112 drain_ready(&mut resampler, &mut dst);
2113 let delay = resampler.delay();
2114 assert!(delay > 0, "the filter has to be holding something");
2115
2116 crate::fault_subprocess::cap_ffmpeg_allocations(1);
2117 let refused = resampler.send_frame(&frame);
2118 crate::fault_subprocess::uncap_ffmpeg_allocations();
2119
2120 assert!(
2121 refused.is_err(),
2122 "an allocator that refuses everything must not look like success",
2123 );
2124 assert_eq!(
2125 resampler.delay(),
2126 delay,
2127 "the frame went into the filter anyway",
2128 );
2129 assert!(
2130 matches!(resampler.receive_frame(&mut dst), Ok(Received::NeedsInput)),
2131 "a failed send left output ready",
2132 );
2133
2134 // And the session is still a session: the same frame converts.
2135 crate::accepted(resampler.send_frame(&frame), "the failure cost nothing");
2136 assert!(matches!(
2137 resampler.receive_frame(&mut dst),
2138 Ok(Received::Frame)
2139 ));
2140 },
2141 );
2142 }
2143
2144 #[test]
2145 fn an_allocation_fault_while_draining_keeps_the_tail() {
2146 // The same property one call along, where the samples at risk are
2147 // the ones already inside the filter: a drain that fails must leave
2148 // the tail where it was, not turn it into an error.
2149 crate::fault_subprocess::in_subprocess(
2150 "resampler::tests::an_allocation_fault_while_draining_keeps_the_tail",
2151 || {
2152 let mut resampler = stereo_to_mono();
2153 let frame = stereo_frame(4_800);
2154 let mut dst = crate::boundary::empty_owned_audio_frame();
2155 for _ in 0..3 {
2156 crate::accepted(resampler.send_frame(&frame), "send_frame");
2157 drain_ready(&mut resampler, &mut dst);
2158 }
2159 crate::accepted(resampler.send_eof(), "eof");
2160 let tail = resampler.delay();
2161 assert!(tail > 0, "there has to be a tail to lose");
2162
2163 crate::fault_subprocess::cap_ffmpeg_allocations(1);
2164 let refused = resampler.receive_frame(&mut dst);
2165 crate::fault_subprocess::uncap_ffmpeg_allocations();
2166
2167 let refused = refused.expect_err("the drain cannot have succeeded");
2168 // No arm of this enum can say "send me more input" any more —
2169 // that answer left the error type — so an allocation failure has
2170 // nowhere to be mistaken for one.
2171 assert!(
2172 matches!(
2173 refused,
2174 ResampleError::Resample(_) | ResampleError::OutputBuffer(_)
2175 ),
2176 "an allocation failure surfaced as something else: {refused:?}",
2177 );
2178 assert_eq!(
2179 resampler.delay(),
2180 tail,
2181 "the tail was consumed by a drain that failed",
2182 );
2183
2184 // And it is still drainable, which is the whole point.
2185 assert_eq!(
2186 resampler
2187 .receive_frame(&mut dst)
2188 .expect("the tail survived the failure"),
2189 Received::Frame,
2190 );
2191 },
2192 );
2193 }
2194
2195 #[test]
2196 fn the_sample_format_table_round_trips() {
2197 for format in [
2198 SampleFormat::U8,
2199 SampleFormat::S16,
2200 SampleFormat::S32,
2201 SampleFormat::S64,
2202 SampleFormat::FLT,
2203 SampleFormat::DBL,
2204 SampleFormat::U8P,
2205 SampleFormat::S16P,
2206 SampleFormat::S32P,
2207 SampleFormat::S64P,
2208 SampleFormat::FLTP,
2209 SampleFormat::DBLP,
2210 ] {
2211 let ffmpeg = format.to_ffmpeg().expect("a named format");
2212 assert_eq!(
2213 SampleFormat::from_ffmpeg(ffmpeg),
2214 format,
2215 "{format:?} does not survive the round trip",
2216 );
2217 assert_eq!(ffmpeg.is_planar(), format.is_planar());
2218 }
2219 assert!(SampleFormat::NONE.to_ffmpeg().is_none());
2220 assert!(SampleFormat::from_raw(9999).to_ffmpeg().is_none());
2221 }
2222
2223 #[test]
2224 fn a_mask_rebuilds_the_layout_it_names() {
2225 let stereo = layout_from_mask(ChannelLayout::STEREO.bits());
2226 assert_eq!(stereo.channels(), 2);
2227 assert_eq!(stereo.bits(), ChannelLayout::STEREO.bits());
2228
2229 let five_one = layout_from_mask(ChannelLayout::_5POINT1.bits());
2230 assert_eq!(five_one.channels(), 6);
2231 assert_eq!(
2232 five_one.bits(),
2233 ChannelLayout::_5POINT1.bits(),
2234 "the side-vs-back distinction is exactly what a default layout would lose",
2235 );
2236 }
2237
2238 #[test]
2239 fn plane_geometry_follows_packed_versus_planar() {
2240 use ffmpeg_next::format::sample::Type;
2241 // Packed: one plane holding every channel.
2242 assert_eq!(
2243 plane_bytes(Sample::I16(Type::Packed), 1024, 2),
2244 Some(1024 * 2 * 2)
2245 );
2246 // Planar: one plane per channel, so the count does not multiply in.
2247 assert_eq!(
2248 plane_bytes(Sample::I16(Type::Planar), 1024, 2),
2249 Some(1024 * 2)
2250 );
2251 assert_eq!(
2252 plane_bytes(Sample::F32(Type::Planar), 1024, 6),
2253 Some(1024 * 4)
2254 );
2255 // A sample count whose byte size does not fit is not a size. This
2256 // is the arithmetic that used to run before the allocation it
2257 // feeds, and it wrapped.
2258 assert_eq!(
2259 plane_bytes(Sample::F32(Type::Packed), usize::MAX / 2, 8),
2260 None,
2261 "an overflowing plane size is refused, not wrapped",
2262 );
2263 }
2264
2265 #[test]
2266 fn the_target_timebase_is_one_tick_per_output_sample() {
2267 let spec = ResampleSpec::new(
2268 16_000,
2269 Sample::I16(ffmpeg_next::format::sample::Type::Packed),
2270 ChannelLayout::MONO,
2271 );
2272 let tb = spec.timebase();
2273 assert_eq!((tb.num(), tb.den().get()), (1, 16_000));
2274 }
2275}