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 owned codec ticket a
99 /// [`crate::FfmpegDemuxer`] track row carries
100 /// (`track.extra().ticket()`) — the "source from `TrackInfo`" path.
101 ///
102 /// Reads the ticket directly rather than rebuilding an
103 /// `AVCodecParameters` to read three fields out of: the seats are
104 /// already owned, and a rebuild here would allocate `extradata` a
105 /// resampler has no use for.
106 ///
107 /// Returns `None` for a non-audio track, for one whose declared
108 /// sample format is `AV_SAMPLE_FMT_NONE` (a codec whose format is
109 /// only known once its decoder opens), and for a custom or ambisonic
110 /// channel layout — see [`Self::from_decoder`] for the first case and
111 /// the note on [`unspecified_layout`] for the last.
112 pub fn from_ticket(ticket: &crate::ticket::CodecTicket) -> Option<Self> {
113 if !crate::boundary::media_kind_from_raw(ticket.codec_type()).is_audio() {
114 return None;
115 }
116 let rate = ticket.sample_rate().max(0) as u32;
117 if rate == 0 {
118 return None;
119 }
120 let format = SampleFormat::from_raw(ticket.format()).to_ffmpeg()?;
121 // Through the same decision [`layout_from_raw`] makes — the same
122 // function, in fact, so the two roads cannot drift into admitting
123 // different layouts from the same file.
124 let owned = ticket.ch_layout();
125 let layout = layout_from_parts(owned.order(), owned.channels(), owned.mask())?;
126 Some(Self::new(rate, format, layout))
127 }
128
129 /// [`Self::from_ticket`], over a live `AVCodecParameters` — the road
130 /// for a caller holding one straight off `stream.parameters()`.
131 ///
132 /// Returns `None` in every case [`Self::from_ticket`] does, and also
133 /// for null-backed parameters: `Parameters`' safe constructors hand
134 /// one back when FFmpeg's allocation failed and report nothing, so a
135 /// caller can arrive here holding one without ever having been told.
136 /// Parameters that were never allocated describe no audio, which
137 /// this function already has a word for.
138 pub fn from_parameters(parameters: &Parameters) -> Option<Self> {
139 // Before `medium()`, which dereferences the pointer inside
140 // ffmpeg-next. `Parameters`' safe constructors hand back a
141 // null-backed value when FFmpeg's allocation failed and report
142 // nothing, so a caller can arrive here holding one without ever
143 // having been told. Parameters that were never allocated describe
144 // no audio, which this function already has a word for.
145 // SAFETY: reading the pointer without dereferencing it.
146 if unsafe { parameters.as_ptr() }.is_null() {
147 return None;
148 }
149 if !crate::boundary::media_kind_of(parameters).is_audio() {
150 return None;
151 }
152 // SAFETY: `parameters` keeps the `AVCodecParameters` live; every
153 // read below goes through the raw pointer and none of them
154 // materialises a bindgen enum out of foreign memory.
155 let par = unsafe { parameters.as_ptr() };
156 let rate = unsafe { (*par).sample_rate }.max(0) as u32;
157 if rate == 0 {
158 return None;
159 }
160 let format = SampleFormat::from_raw(unsafe { (*par).format }).to_ffmpeg()?;
161 let layout = unsafe { layout_from_raw(addr_of!((*par).ch_layout)) }?;
162 Some(Self::new(rate, format, layout))
163 }
164
165 /// The spec an opened decoder will actually produce — its rate,
166 /// sample format and channel layout, straight off the codec context.
167 ///
168 /// Reach it through
169 /// [`FfmpegAudioStreamDecoder::inner`](crate::FfmpegAudioStreamDecoder::inner).
170 /// `None` on a custom or ambisonic layout, and on a context whose
171 /// sample format is still unset (a decoder that has not been opened).
172 pub fn from_decoder(decoder: &ffmpeg_next::decoder::Audio) -> Option<Self> {
173 // SAFETY: `decoder` keeps the `AVCodecContext` live. `sample_fmt`
174 // is read as the raw integer it is rather than through
175 // `decoder.format()`, which would construct an `AVSampleFormat`
176 // out of foreign memory.
177 let ctx = unsafe { decoder.as_ptr() };
178 // Same reason as `from_parameters`: `codec::Context::new()` is a
179 // safe constructor over an unchecked `avcodec_alloc_context3`, so a
180 // decoder can be null-backed without anyone having been told.
181 if ctx.is_null() {
182 return None;
183 }
184 let format =
185 SampleFormat::from_raw(unsafe { read_unaligned(addr_of!((*ctx).sample_fmt).cast::<i32>()) })
186 .to_ffmpeg()?;
187 let rate = unsafe { (*ctx).sample_rate }.max(0) as u32;
188 if rate == 0 {
189 return None;
190 }
191 let layout = unsafe { layout_from_raw(addr_of!((*ctx).ch_layout)) }?;
192 Some(Self::new(rate, format, layout))
193 }
194
195 /// A layout that names a channel *count* and nothing else —
196 /// `AV_CHANNEL_ORDER_UNSPEC`.
197 ///
198 /// Not a degenerate case: a WAV file without a `WAVE_FORMAT_EXTENSIBLE`
199 /// channel mask genuinely declares no layout, and FFmpeg faithfully
200 /// reports it as unspecified in the codec parameters, in the codec
201 /// context, and on every decoded frame. Substituting a default layout
202 /// would make the source spec disagree with the frames it is supposed
203 /// to describe, and every `send_frame` would be refused as a
204 /// mid-stream change. `swr` accepts an unspecified layout at either
205 /// end and maps the channels positionally.
206 #[inline]
207 pub fn unspecified_layout(channels: i32) -> ChannelLayout {
208 // SAFETY: a zeroed `AVChannelLayout` is a valid value — `order`
209 // reads as `AV_CHANNEL_ORDER_UNSPEC`, the zero discriminant, and
210 // the union is documented as unused for that order.
211 unsafe {
212 let mut layout: ffmpeg_next::ffi::AVChannelLayout = std::mem::zeroed();
213 layout.nb_channels = channels.max(0);
214 ChannelLayout(layout)
215 }
216 }
217
218 /// Sample rate in Hz.
219 #[inline]
220 pub const fn rate(&self) -> u32 {
221 self.rate
222 }
223 /// Sample format.
224 #[inline]
225 pub const fn format(&self) -> Sample {
226 self.format
227 }
228 /// Channel layout.
229 #[inline]
230 pub const fn layout(&self) -> ChannelLayout {
231 self.layout
232 }
233 /// Channel count, from the layout.
234 #[inline]
235 pub fn channels(&self) -> i32 {
236 self.layout.channels()
237 }
238
239 /// The timebase output frames carry — one tick per output sample.
240 fn timebase(&self) -> Timebase {
241 Timebase::new(
242 1,
243 std::num::NonZeroI32::new(self.rate.min(i32::MAX as u32) as i32).unwrap_or(
244 // A zero-rate spec never reaches here: `new` is the only way in
245 // and every caller of it names a real rate. Falling back to
246 // one tick per second keeps the arithmetic total rather than
247 // panicking on a value that cannot occur.
248 std::num::NonZeroI32::new(1).expect("1 is non-zero"),
249 ),
250 )
251 }
252}
253
254/// `mediadecode::resampler::AudioResampler` impl wrapping
255/// `swresample`.
256///
257/// Construction is [`Self::new`], off the trait, taking both specs —
258/// see the trait's own documentation for why the target can never be a
259/// constant.
260pub struct CarrierResampler<C: crate::FfmpegCarrier> {
261 ctx: resampling::Context,
262 source: ResampleSpec,
263 target: ResampleSpec,
264 /// The source spec restated in the vocabulary a decoded `AudioFrame`
265 /// carries. The mid-stream check compares against these, not against
266 /// FFmpeg's dialect, so it never has to translate a frame.
267 source_format: SampleFormat,
268 source_layout: ChannelLayoutDescription,
269 /// The target spec in the vocabulary an output frame carries,
270 /// computed once at construction. Assembling a converted frame after
271 /// `swr` has run must not have to ask FFmpeg anything, because asking
272 /// can fail — see [`FfmpegResampler::prepare_output`].
273 target_format: SampleFormat,
274 target_layout: ChannelLayoutDescription,
275 /// The layouts `swr` is really configured with — see
276 /// [`initialized_layout`]. Every `AVFrame` this type stages or
277 /// allocates carries these, not the declared ones.
278 staged_source_layout: ChannelLayout,
279 staged_target_layout: ChannelLayout,
280 target_timebase: Timebase,
281 ready: VecDeque<Frame<C>>,
282 /// Next output timestamp, in target-rate ticks. `None` until the
283 /// first input frame anchors it.
284 next_pts: Option<i64>,
285 eof: bool,
286 /// `true` once the post-EOF tail has been drained to its end.
287 ///
288 /// **The terminal answer has to be terminal.** Without this latch,
289 /// every poll past the end re-enters the flush road — allocating an
290 /// output frame and asking `swr` again — because `swr_get_delay` can
291 /// keep reporting residual samples that `swr_convert_frame` will
292 /// never emit (observed: 16 samples left standing after the tail is
293 /// genuinely exhausted). That is harmless while allocation succeeds
294 /// and wrong when it does not: a session that has already answered
295 /// `Ended` would start answering with an allocation error instead,
296 /// turning a settled protocol state back into a fault. The latch
297 /// makes the end cheap and unconditional. `flush` clears it with the
298 /// rest of the session.
299 drained: bool,
300 /// What one converted frame may cost. See
301 /// [`Self::check_output_bytes`] for why a resampler needs a ceiling
302 /// of its own even when its input already had one.
303 limits: FrameLimits,
304 /// The lane. Zero-sized: it selects how a produced plane is carried
305 /// — shared out of the output `AVFrame` or copied out of it — and
306 /// nothing else about the conversion.
307 _carrier: core::marker::PhantomData<C>,
308}
309
310impl<C: crate::FfmpegCarrier + crate::CarrierOps> CarrierResampler<C> {
311 /// Opens a resampler between two explicit specs.
312 ///
313 /// Both are required and neither is inferred. The source is what the
314 /// decoder will hand over — read it off the track
315 /// ([`ResampleSpec::from_parameters`]) or off the opened decoder
316 /// ([`ResampleSpec::from_decoder`]). The target is the caller's, and
317 /// is options: 16 kHz mono for a speech model, 48 kHz for an
318 /// audio-event one, both from the same track.
319 ///
320 /// # The choke point
321 ///
322 /// [`ResampleSpec::new`] is `const` and total, so this is where both
323 /// ends are checked — every construction route (`from_parameters`,
324 /// `from_decoder`, the public constructor) passes through here, and
325 /// nothing hazardous reaches `swr` or a staged `AVFrame` behind it:
326 ///
327 /// - a rate of zero, or one past `c_int`
328 /// ([`ResampleError::UnsupportedRate`]);
329 /// - `AV_SAMPLE_FMT_NONE` ([`ResampleError::UnsupportedFormat`]);
330 /// - a channel layout that is neither native nor unspecified, or one
331 /// naming no channels ([`ResampleError::UnsupportedLayout`]).
332 ///
333 /// `limits` bounds what one **converted** frame may cost.
334 ///
335 /// [`FrameLimits`] rather than a seat of this seam's own: the
336 /// quantity is bytes of one produced audio frame, which is exactly
337 /// what [`FrameLimits::max_frame_bytes`] means everywhere else in
338 /// this crate, and one number for "what a frame may cost" is worth
339 /// more than a second vocabulary. [`FrameLimits::max_pixels`] is
340 /// unused here, as it is on the audio decode path, and for the same
341 /// reason: audio has no pixels.
342 pub(crate) fn new_impl(
343 source: ResampleSpec,
344 target: ResampleSpec,
345 limits: FrameLimits,
346 ) -> Result<Self, ResampleError> {
347 check_spec(&source, SpecEnd::Source)?;
348 check_spec(&target, SpecEnd::Target)?;
349
350 // The layouts `swr` is really configured with, resolved *before*
351 // the pair is judged — because the conversion that will run is
352 // between these two, not between the two that were declared. An
353 // unspecified layout becomes FFmpeg's default for its channel count
354 // (twenty-four unspecified channels are 22.2), so judging the
355 // declared pair let exactly the routing the explicit 22.2 refusal
356 // blocks walk in through the unspecified door.
357 let staged_source_layout = initialized_layout(source.layout);
358 let staged_target_layout = initialized_layout(target.layout);
359 check_pair(&staged_source_layout, &staged_target_layout)?;
360 let ctx = open_context(&source, &target, staged_source_layout, staged_target_layout)?;
361
362 let source_format = SampleFormat::from_ffmpeg(source.format);
363 let target_format = SampleFormat::from_ffmpeg(target.format);
364 // SAFETY: the layout is a live `ChannelLayout` owned by this scope.
365 let target_layout =
366 crate::channel_layout::channel_layout_description_from_ffmpeg(&staged_target_layout);
367 // SAFETY: the layout is a live `ChannelLayout` owned by `source`
368 // for the duration of this call.
369 let source_layout =
370 crate::channel_layout::channel_layout_description_from_ffmpeg(&source.layout);
371 let target_timebase = target.timebase();
372
373 Ok(Self {
374 ctx,
375 source,
376 target,
377 source_format,
378 source_layout,
379 target_format,
380 target_layout,
381 staged_source_layout,
382 staged_target_layout,
383 target_timebase,
384 ready: VecDeque::new(),
385 next_pts: None,
386 eof: false,
387 drained: false,
388 limits,
389 _carrier: core::marker::PhantomData,
390 })
391 }
392
393 /// The spec frames must arrive in.
394 #[inline]
395 pub(crate) const fn source_impl(&self) -> &ResampleSpec {
396 &self.source
397 }
398
399 /// The spec frames leave in.
400 #[inline]
401 pub(crate) const fn target_impl(&self) -> &ResampleSpec {
402 &self.target
403 }
404
405 /// Borrows the wrapped `swr` context.
406 #[inline]
407 pub(crate) const fn inner_impl(&self) -> &resampling::Context {
408 &self.ctx
409 }
410
411 /// Samples still inside the delay line, counted at the output rate.
412 #[inline]
413 pub(crate) fn delay_impl(&self) -> i64 {
414 self.ctx.delay().map_or(0, |d| d.output.max(0))
415 }
416
417 /// Refuses a frame whose shape is not the source spec.
418 fn check_source(&self, frame: &Frame<C>) -> Result<(), ResampleError> {
419 if frame.sample_rate() != self.source.rate
420 || *frame.sample_format() != self.source_format
421 || *frame.channel_layout() != self.source_layout
422 {
423 return Err(ResampleError::SourceChanged(SourceChanged::new(
424 self.source.rate,
425 self.source_format,
426 frame.sample_rate(),
427 *frame.sample_format(),
428 )));
429 }
430 Ok(())
431 }
432
433 /// Where a frame's timestamp lands on the output timeline, or `None`
434 /// when it carries none.
435 ///
436 /// Rescaled with the **checked** rung, and before anything is staged.
437 /// `Timestamp::rescale_to` saturates, and both ends of that clamp are
438 /// wrong here: a positive one reaches the counted timeline's checked
439 /// addition only after `swr` has consumed the input, leaving a
440 /// session no caller can retry; a negative one lands on `i64::MIN`,
441 /// which *is* `AV_NOPTS_VALUE`, so the conversion back reads the
442 /// frame as carrying no timestamp at all and an extreme timestamp is
443 /// silently erased. A timestamp that does not fit the output timeline
444 /// is refused by name, with the resampler untouched.
445 fn anchor_of(&self, frame: &Frame<C>) -> Result<Option<i64>, ResampleError> {
446 let Some(timestamp) = frame.pts() else {
447 return Ok(None);
448 };
449 let ticks = timestamp.pts();
450 let out_of_range = || ResampleError::TimestampOutOfRange(TimestampOutOfRange::new(ticks));
451 // `AV_NOPTS_VALUE` is a sentinel, not a time. A frame carrying it
452 // as a value says something contradictory, and anchoring on it
453 // would produce output frames that report no timestamp.
454 if ticks == AV_NOPTS_VALUE {
455 return Err(out_of_range());
456 }
457 let rescaled = timestamp
458 .timebase()
459 .checked_rescale(ticks, self.target_timebase)
460 .ok_or_else(out_of_range)?;
461 if rescaled == AV_NOPTS_VALUE {
462 return Err(out_of_range());
463 }
464 Ok(Some(rescaled))
465 }
466
467 /// Stages a decoded frame as an `AVFrame` swr can read.
468 ///
469 /// Geometry is settled **before** anything is allocated. A frame's
470 /// header is a claim, not a fact: `nb_samples` comes from the same
471 /// foreign memory as the planes it describes, and sizing an
472 /// allocation off it first would let a forged frame with a
473 /// twelve-byte plane ask for tens of gigabytes on its way to being
474 /// refused.
475 fn stage_input(&self, frame: &Frame<C>) -> Result<frame::Audio, ResampleError> {
476 let samples = frame.nb_samples() as usize;
477 let channels = self.source.channels();
478
479 // What the *format* requires, not what the allocated frame reports
480 // — the frame does not exist yet.
481 let planes = if self.source.format.is_planar() {
482 // `check_spec` proved this positive at construction.
483 channels as usize
484 } else {
485 1
486 };
487 let found = frame.plane_count() as usize;
488 if planes > found {
489 return Err(ResampleError::PlaneCount(PlaneCount::new(planes, found)));
490 }
491 let bytes = plane_bytes(self.source.format, samples, channels)
492 .ok_or(ResampleError::SampleCount(SampleCount::new(samples)))?;
493 for plane in frame.planes().iter().take(planes) {
494 let src = plane.data_ref().as_ref();
495 if src.len() < bytes {
496 return Err(ResampleError::PlaneCount(PlaneCount::new(bytes, src.len())));
497 }
498 }
499
500 // Only now, with every plane proved long enough for the sample
501 // count that sizes this allocation.
502 let mut input = new_audio_frame(
503 self.source.format,
504 samples,
505 self.source.rate,
506 self.staged_source_layout,
507 )?;
508 // What the allocation really produced. `data_mut` panics past its
509 // own plane count, and this crate does not put a panic on a path
510 // that reads foreign geometry.
511 let staged = input.planes();
512 if staged < planes {
513 return Err(ResampleError::PlaneCount(PlaneCount::new(planes, staged)));
514 }
515 for (index, plane) in frame.planes().iter().take(planes).enumerate() {
516 let src = plane.data_ref().as_ref();
517 let dst = input.data_mut(index);
518 if dst.len() < bytes {
519 return Err(ResampleError::PlaneCount(PlaneCount::new(bytes, dst.len())));
520 }
521 dst[..bytes].copy_from_slice(&src[..bytes]);
522 }
523 Ok(input)
524 }
525
526 /// The most samples the next conversion could produce: the delay
527 /// line's contents plus `in_samples` of new input, rescaled to the
528 /// output rate and rounded up.
529 ///
530 /// Separate from the allocation because it is also the preflight the
531 /// output timeline is checked against — *before* `swr` consumes
532 /// anything, so a refusal leaves the session where a caller can retry
533 /// it.
534 fn output_capacity(&self, in_samples: i64) -> Result<usize, ResampleError> {
535 let delay_in = self.ctx.delay().map_or(0, |d| d.input.max(0));
536 let total = delay_in.saturating_add(in_samples).max(0) as i128;
537 let scaled = (total * i128::from(self.target.rate) + i128::from(self.source.rate) - 1)
538 / i128::from(self.source.rate).max(1);
539 // One extra sample of headroom: swr rounds its own accounting, and
540 // an output frame one short would silently push the remainder into
541 // the internal FIFO where the pts accounting cannot see it until
542 // the next call.
543 let samples = scaled + 1;
544 // `av_frame_get_buffer` takes the count as a `c_int`. A request
545 // past that is refused by name rather than clamped: a silently
546 // shortened output frame is a stream that loses samples.
547 if samples > i128::from(i32::MAX) {
548 // Saturating only for a count past `usize` itself, which no
549 // machine could hold either way.
550 return Err(ResampleError::SampleCount(SampleCount::new(
551 usize::try_from(samples).unwrap_or(usize::MAX),
552 )));
553 }
554 Ok(samples.max(1) as usize)
555 }
556
557 /// Refuses a conversion whose output would not fit the frame ceiling.
558 ///
559 /// # Why a resampler needs one even though its input had one
560 ///
561 /// [`Self::output_capacity`] bounds the output **sample count**, and
562 /// only at `i32::MAX` — the structural limit of
563 /// `av_frame_get_buffer`. Nothing in it bounds the *bytes*, and the
564 /// two are related by a ratio the caller does not control: the
565 /// capacity is `input_samples × target_rate / source_rate`, and a
566 /// source spec read off an untrusted container can say 1 Hz. One
567 /// second of 1 Hz mono input converted to 48 kHz stereo `f32` is
568 /// 384 KiB from 4 bytes — and the same input against a 1 Hz source
569 /// claim and a 192 kHz target is multi-gigabyte. The frame that
570 /// arrived was within *its* ceiling; the frame that leaves need not
571 /// be, so it gets its own judgement.
572 ///
573 /// Both allocations are covered: `av_frame_get_buffer`'s, and the
574 /// [`FfmpegBytes`] copy [`Self::finish_output`] makes from it.
575 fn check_output_bytes(&self, capacity: usize, channels: i32) -> Result<(), ResampleError> {
576 // **Priced as the allocator prices it, not as the samples weigh.**
577 // This used to multiply the tight plane length by the plane count,
578 // which is the payload's arithmetic and not
579 // `av_frame_get_buffer`'s: a one-sample eight-channel planar `s16`
580 // frame is 16 bytes of samples and a **768-byte** allocation,
581 // because every plane is aligned and padded on its own. A 16-byte
582 // ceiling admitted it.
583 //
584 // The overhead is under one percent on any frame big enough to
585 // care about, which is exactly why this went unseen — see
586 // [`crate::footprint`] for the measured table and for the rule the
587 // whole crate now keeps: a judge must dominate the allocator's
588 // arithmetic, not the payload's.
589 let bytes = crate::footprint::audio_frame_bytes(
590 ffmpeg_next::ffi::AVSampleFormat::from(self.target.format) as libc::c_int,
591 capacity,
592 channels.max(0) as usize,
593 )
594 .ok_or(ResampleError::OutputTooLarge(OutputTooLarge::new(
595 usize::MAX,
596 self.limits.max_frame_bytes(),
597 )))?;
598 if bytes > self.limits.max_frame_bytes() {
599 return Err(ResampleError::OutputTooLarge(OutputTooLarge::new(
600 bytes,
601 self.limits.max_frame_bytes(),
602 )));
603 }
604 Ok(())
605 }
606
607 /// Refuses a conversion whose output could not be labelled: the
608 /// timeline plus everything this call might produce has to stay
609 /// inside `i64`.
610 ///
611 /// Asked before `swr` sees a sample, like everything else that can
612 /// fail. [`Self::finish_output`] performs the same addition against
613 /// the count actually produced, which cannot exceed the capacity
614 /// checked here — so once this passes, that one cannot fail.
615 fn check_timeline(&self, anchor: Option<i64>, capacity: usize) -> Result<(), ResampleError> {
616 let pts = self.next_pts.or(anchor).unwrap_or(0);
617 let samples = capacity as i64;
618 if pts.checked_add(samples).is_none() {
619 return Err(ResampleError::TimestampOverflow(TimestampOverflow::new(
620 pts, samples,
621 )));
622 }
623 Ok(())
624 }
625
626 /// Allocates the output frame **and proves every plane the converted
627 /// frame will be read out of**, before `swr` is allowed to touch a
628 /// sample.
629 ///
630 /// This is the shape the whole seam is built around. Anything
631 /// fallible that runs *after* `swr_convert_frame` has consumed input
632 /// leaves a session no caller can act on: retrying feeds the same
633 /// samples twice, continuing loses them, and the delay line has moved
634 /// either way. The failure kept relocating — the timestamp addition,
635 /// the tail drain, then the output wrapping — so the fix is not
636 /// another check in another place but an ordering that leaves nothing
637 /// on the far side: the frame, every plane pointer it will be read
638 /// through, and the queue slot are all taken here, where failing
639 /// costs nothing but an error.
640 ///
641 /// **What 0.9 changed, and what it did not.** Through 0.8 this
642 /// function also *acquired* one `AVBufferRef` view per plane, because
643 /// wrapping a plane could fail and so had to happen on this side of
644 /// the conversion; [`Self::finish_output`] then narrowed each view to
645 /// what `swr` produced. The amputation removes the views — the output
646 /// planes are copied out afterwards instead — and with them the
647 /// failure that forced the acquisition to be early. What stays early
648 /// is the *proof*: every plane pointer is checked non-null and
649 /// checked to address `plane_len` bytes inside one of the frame's own
650 /// buffers here, so the copy on the far side has nothing left to
651 /// judge and [`Self::finish_output`] remains infallible.
652 fn prepare_output(&self, capacity: usize) -> Result<PreparedOutput<C>, ResampleError> {
653 let channels = self.target.channels();
654 let plane_count = if self.target.format.is_planar() {
655 // `check_spec` proved this positive at construction.
656 channels as usize
657 } else {
658 1
659 };
660 let plane_len = plane_bytes(self.target.format, capacity, channels)
661 .ok_or(ResampleError::SampleCount(SampleCount::new(capacity)))?;
662 // Linear in the sample count, which is what lets the post-run
663 // trim be a multiplication rather than another fallible call.
664 let per_sample = plane_bytes(self.target.format, 1, channels)
665 .ok_or(ResampleError::SampleCount(SampleCount::new(1)))?;
666
667 // **The byte ceiling, before `av_frame_get_buffer` and before the
668 // carrier copy that follows it.** Measured first, allocated second:
669 // the three quantities above are arithmetic over the target spec
670 // and cost nothing, so there is no reason for the frame to exist
671 // before the answer does.
672 self.check_output_bytes(capacity, channels)?;
673
674 let frame = new_audio_frame(
675 self.target.format,
676 capacity,
677 self.target.rate,
678 self.staged_target_layout,
679 )?;
680 if frame.planes() < plane_count {
681 return Err(ResampleError::PlaneCount(PlaneCount::new(
682 plane_count,
683 frame.planes(),
684 )));
685 }
686
687 let mut reserved: [Option<C::Reserved>; 8] = core::array::from_fn(|_| None);
688 for (index, slot) in reserved.iter_mut().enumerate().take(plane_count) {
689 // SAFETY: `frame` owns a live `AVFrame` this call just
690 // allocated; `data` is a public field and `plane_count` is
691 // within the eight slots `data` has.
692 let data_ptr = unsafe { (*frame.as_ptr()).data[index] };
693 if data_ptr.is_null() {
694 return Err(ResampleError::OutputBuffer(OutputBuffer::new(index)));
695 }
696 // SAFETY: the frame is live, and the helper only reads `buf[]`'s
697 // ranges to find the one containing `data_ptr`. Proving it here
698 // is what lets the carry on the far side of the conversion be
699 // unconditional — a copy on the owned lane, a view on the other,
700 // and neither has a proof left to make.
701 let backing =
702 unsafe { crate::convert::find_audio_backing_buffer(frame.as_ptr(), data_ptr, plane_len) }
703 .ok_or(ResampleError::OutputBuffer(OutputBuffer::new(index)))?;
704 // The carrier's claim on the plane, taken **here** — before `swr`
705 // consumes anything. See [`FfmpegCarrier::reserve`].
706 //
707 // SAFETY: `backing` is a live buffer of this frame's, proved just
708 // above to cover `plane_len` bytes from `data_ptr`; the frame is
709 // moved into `PreparedOutput` below and so outlives the commit.
710 let offset = unsafe { (data_ptr as usize).wrapping_sub((*backing).data as usize) };
711 // SAFETY: as above — the extent lies inside `backing`.
712 *slot = Some(
713 unsafe { C::reserve(backing, offset, plane_len) }
714 .ok_or(ResampleError::OutputBuffer(OutputBuffer::new(index)))?,
715 );
716 }
717
718 Ok(PreparedOutput {
719 frame,
720 reserved,
721 plane_count,
722 plane_len,
723 per_sample,
724 })
725 }
726
727 /// Turns a converted frame into a `mediadecode` one. **Infallible**,
728 /// by construction: every check it could have made was made in
729 /// [`Self::prepare_output`], and everything left here is arithmetic
730 /// over values this type owns plus a copy that cannot be refused.
731 ///
732 /// `None` when the conversion produced nothing — the delay line
733 /// swallowed the input, which is ordinary and not a failure.
734 fn finish_output(&mut self, mut prepared: PreparedOutput<C>) -> Option<Frame<C>> {
735 let produced = prepared.frame.samples();
736 if produced == 0 {
737 return None;
738 }
739 let pts = self.next_pts.unwrap_or(0);
740 // `check_timeline` ran before `swr` did, against a capacity that is
741 // never smaller than what came out, so this addition cannot leave
742 // `i64`. It is stated rather than checked because a check here
743 // would be an error path on the wrong side of the conversion —
744 // exactly what this design exists to remove.
745 debug_assert!(
746 pts.checked_add(produced as i64).is_some(),
747 "the timeline was preflighted against a capacity >= produced",
748 );
749 self.next_pts = Some(pts.saturating_add(produced as i64));
750
751 let bytes = prepared
752 .per_sample
753 .saturating_mul(produced)
754 .min(prepared.plane_len);
755 let plane_count = prepared.plane_count;
756 let mut planes: [Plane<C::Buffer>; 8] = core::array::from_fn(|_| Plane::new(C::empty(), 0));
757 for (index, slot) in planes.iter_mut().enumerate().take(plane_count) {
758 // Present for every index below `plane_count`: `prepare_output`
759 // fills exactly that many and returns an error otherwise.
760 let Some(reserved) = prepared.reserved[index].take() else {
761 continue;
762 };
763 // **The valid prefix, not the plane.** `plane_len` is what
764 // capacity was allocated for; `bytes` is what `swr` produced.
765 // Both lanes stop at the latter, for the same reason the decode
766 // road does: the tail is allocator memory nothing wrote, and a
767 // carrier's span is a span a consumer may read.
768 //
769 // SAFETY: the reservation covers `plane_len` bytes inside one of
770 // `frame`'s own buffers, and `bytes <= plane_len` by the `min`
771 // above. `swr_convert_frame` has written those bytes and does not
772 // replace the buffers; `frame` has been owned by `prepared` — and
773 // so kept alive — across the whole conversion. A view committed
774 // here outlives `prepared` by refcount, and the writing is over
775 // before the sharing begins: this resampler allocates a fresh
776 // output frame per conversion, so nothing ever writes into a
777 // buffer a delivered frame is reading.
778 *slot = Plane::new(unsafe { C::commit(reserved, bytes) }, bytes as u32);
779 }
780
781 Some(
782 AudioFrame::new(
783 self.target.rate,
784 produced as u32,
785 // Exact, not clipped: `check_spec` refused every spec outside
786 // `1..=MAX_FRAME_CHANNELS` before this resampler existed.
787 self.target.channels() as u8,
788 self.target_format,
789 self.target_layout.clone(),
790 planes,
791 plane_count as u8,
792 AudioFrameExtra::default(),
793 )
794 .with_pts(Some(Timestamp::new(pts, self.target_timebase)))
795 .with_duration(Some(Timestamp::new(produced as i64, self.target_timebase))),
796 )
797 }
798}
799
800/// Everything a converted frame needs, acquired before the conversion
801/// runs. See [`FfmpegResampler::prepare_output`].
802struct PreparedOutput<C: crate::FfmpegCarrier + crate::CarrierOps> {
803 /// The output `AVFrame`. Owning it here is what keeps the plane
804 /// pointers below valid across the conversion — moving this struct
805 /// moves a pointer to the `AVFrame`, never the `AVFrame` or the
806 /// buffers it addresses.
807 frame: frame::Audio,
808 /// One carrier claim per populated plane, taken before the
809 /// conversion ran and settled at its true length after — which is
810 /// what keeps this struct's whole reason for existing intact on the
811 /// view lane too. `None` for every unpopulated slot.
812 reserved: [Option<C::Reserved>; 8],
813 plane_count: usize,
814 /// Bytes one plane holds at full capacity — the ceiling every trim
815 /// stays under.
816 plane_len: usize,
817 /// Bytes one plane holds per sample.
818 per_sample: usize,
819}
820
821impl<C: crate::FfmpegCarrier + crate::CarrierOps> CarrierResampler<C> {
822 /// **Always [`Sent::Accepted`] when it accepts at all.** The
823 /// converted-frame queue this type keeps is unbounded — every frame a
824 /// conversion produces is built before `swr` is asked and pushed
825 /// straight onto it — so there is no state in which draining first
826 /// would let a submission through that is refused now. A bounded
827 /// implementation of the same face would answer
828 /// [`Sent::MustDrain`] here; this one has nothing to say it about.
829 ///
830 /// [`ResampleError::AfterEof`] stays an error rather than becoming
831 /// that arm, and the line is the same one the decoders draw: a
832 /// resampler that has been told the stream ended will refuse this
833 /// frame however much is drained first, so sending the caller into a
834 /// drain loop would be sending it nowhere. `flush` is the way back,
835 /// and the message says so.
836 pub(crate) fn send_frame_impl(&mut self, frame: &Frame<C>) -> Result<Sent, ResampleError> {
837 if self.eof {
838 return Err(ResampleError::AfterEof);
839 }
840 self.check_source(frame)?;
841 // A frame carrying no samples is a header and nothing else. There
842 // is nothing to convert and nothing to stage: `av_frame_get_buffer`
843 // refuses a zero-sample allocation, so staging one would hand `swr`
844 // an unbacked `AVFrame` for no gain.
845 if frame.nb_samples() == 0 {
846 return Ok(Sent::Accepted);
847 }
848
849 // Nothing below touches the session's state until the conversion
850 // has succeeded. A refused frame must leave the timeline exactly
851 // where it was, or the next good frame inherits the rejected one's
852 // timestamp.
853 let anchor = self.anchor_of(frame)?;
854 let input = self.stage_input(frame)?;
855 let capacity = self.output_capacity(frame.nb_samples() as i64)?;
856 self.check_timeline(anchor, capacity)?;
857 let mut prepared = self.prepare_output(capacity)?;
858 // The last fallible thing before the conversion: room for the frame
859 // it will produce. `push_back` on a full queue allocates, and an
860 // allocation failure there aborts the process rather than
861 // unwinding — so the growth happens here, where it can be an error.
862 self
863 .ready
864 .try_reserve(1)
865 .map_err(|_| ResampleError::QueueAlloc)?;
866
867 // The only mutation. Everything above could fail and cost nothing;
868 // nothing below can fail at all.
869 self
870 .ctx
871 .run(&input, &mut prepared.frame)
872 .map_err(|e| ResampleError::Resample(Error::Ffmpeg(e)))?;
873
874 // The frame is inside the filter now, so the timeline may be
875 // anchored on it. Anchored on *input* rather than on the first
876 // output, because a call that produces nothing but fills the delay
877 // line still fixes where the stream starts.
878 if self.next_pts.is_none() {
879 self.next_pts = anchor;
880 }
881 if let Some(converted) = self.finish_output(prepared) {
882 self.ready.push_back(converted);
883 }
884 Ok(Sent::Accepted)
885 }
886
887 /// **No parked-frame seat here, and none is needed.** The queue holds
888 /// frames that are already built: every fallible step of a
889 /// conversion — the carrier's claim included — happens in
890 /// [`Self::prepare_output`], before `swr` consumes anything, and
891 /// `finish_output` is infallible by construction. There is no
892 /// conversion left to fail after a frame has been taken out of the
893 /// queue, so nothing can be lost between the two. That is the
894 /// property the reserve-then-commit seam was built for, stated where
895 /// the sibling roads state their seats.
896 pub(crate) fn receive_frame_impl(
897 &mut self,
898 dst: &mut Frame<C>,
899 ) -> Result<Received, ResampleError> {
900 if let Some(frame) = self.ready.pop_front() {
901 *dst = frame;
902 return Ok(Received::Frame);
903 }
904 if !self.eof {
905 return Ok(Received::NeedsInput);
906 }
907 if self.drained {
908 return Ok(Received::Ended);
909 }
910 // EOF: drain the conversion tail. Without this every file loses the
911 // tens of milliseconds sitting inside the filter.
912 //
913 // **This is where the two answers used to be one.** Pre-EOF nothing
914 // ready and post-EOF tail exhausted both returned `Again`, so a
915 // caller that did not itself remember whether it had called
916 // `send_eof` could not tell "send more" from "there is no more" —
917 // and a drain loop written against the seam alone spun forever on
918 // a resampler that was already finished.
919 let remaining = self.delay_impl();
920 if remaining <= 0 {
921 self.drained = true;
922 return Ok(Received::Ended);
923 }
924 let capacity = remaining.min(i64::from(i32::MAX)) as usize;
925 // Same discipline as `send_frame`, and for the same reason: the
926 // tail is drained only once the timeline can hold it and every
927 // reference the converted frame needs is already in hand, so a
928 // failure leaves the delay line untouched instead of turning
929 // samples into an error.
930 self.check_timeline(None, capacity)?;
931 let mut prepared = self.prepare_output(capacity)?;
932 self
933 .ctx
934 .flush(&mut prepared.frame)
935 .map_err(|e| ResampleError::Resample(Error::Ffmpeg(e)))?;
936 match self.finish_output(prepared) {
937 Some(frame) => {
938 *dst = frame;
939 Ok(Received::Frame)
940 }
941 // The delay line reported samples and the flush produced none.
942 // Another flush would report the same and produce the same — this
943 // really happens, `swr_get_delay` standing at a residue the
944 // converter will not emit — so this is the end of the tail and
945 // not a pause in it. Saying otherwise is the spin this reform
946 // removes.
947 None => {
948 self.drained = true;
949 Ok(Received::Ended)
950 }
951 }
952 }
953
954 pub(crate) fn send_eof_impl(&mut self) -> Result<Sent, ResampleError> {
955 self.eof = true;
956 Ok(Sent::Accepted)
957 }
958
959 /// Resets the resampler for another stream on the same two specs.
960 ///
961 /// The `swr` context is **rebuilt**, not drained. `swresample` has no
962 /// reset call, and draining it dry cannot be verified from outside: a
963 /// `swr_convert_frame` that makes no progress reports no error, so a
964 /// drain loop that gives up and a drain loop that finished are
965 /// indistinguishable — and a flush that returned `Ok` with the old
966 /// delay line still inside would let one stream's tail contaminate
967 /// the next. A fresh context is the only reset whose success is a
968 /// fact.
969 ///
970 /// The new context is built before the old one is dropped, so a
971 /// failure leaves the resampler exactly as it was: this call either
972 /// resets everything or changes nothing.
973 pub(crate) fn flush_impl(&mut self) -> Result<(), ResampleError> {
974 let ctx = open_context(
975 &self.source,
976 &self.target,
977 self.staged_source_layout,
978 self.staged_target_layout,
979 )?;
980 self.ctx = ctx;
981 self.ready.clear();
982 self.next_pts = None;
983 self.eof = false;
984 self.drained = false;
985 debug_assert_eq!(self.delay_impl(), 0, "a fresh swr context holds nothing");
986 Ok(())
987 }
988}
989
990macro_rules! resampler_lane_face {
991 ($($lane:ty),+ $(,)?) => { $(
992 impl CarrierResampler<$lane> {
993 /// Opens a resampler between two explicit specs. See
994 /// [`CarrierResampler::new_impl`] for the full contract.
995 pub fn new(
996 source: ResampleSpec,
997 target: ResampleSpec,
998 limits: FrameLimits,
999 ) -> Result<Self, ResampleError> {
1000 Self::new_impl(source, target, limits)
1001 }
1002
1003 /// The spec frames must arrive in.
1004 pub const fn source(&self) -> &ResampleSpec {
1005 self.source_impl()
1006 }
1007
1008 /// The spec frames leave in.
1009 pub const fn target(&self) -> &ResampleSpec {
1010 self.target_impl()
1011 }
1012
1013 /// The wrapped `swr` context.
1014 pub const fn inner(&self) -> &resampling::Context {
1015 self.inner_impl()
1016 }
1017
1018 /// Samples still inside the delay line, counted at the output
1019 /// rate.
1020 pub fn delay(&self) -> i64 {
1021 self.delay_impl()
1022 }
1023 }
1024
1025 impl AudioResampler for CarrierResampler<$lane> {
1026 type Adapter = Ffmpeg;
1027 type Buffer = <$lane as crate::FfmpegCarrier>::Buffer;
1028 type Error = ResampleError;
1029
1030 fn send_frame(&mut self, frame: &Frame<$lane>) -> Result<Sent, ResampleError> {
1031 self.send_frame_impl(frame)
1032 }
1033
1034 fn receive_frame(&mut self, dst: &mut Frame<$lane>) -> Result<Received, ResampleError> {
1035 self.receive_frame_impl(dst)
1036 }
1037
1038 fn send_eof(&mut self) -> Result<Sent, ResampleError> {
1039 self.send_eof_impl()
1040 }
1041
1042 fn flush(&mut self) -> Result<(), ResampleError> {
1043 self.flush_impl()
1044 }
1045 }
1046 )+ };
1047}
1048
1049resampler_lane_face!(crate::View, crate::Owned);
1050
1051/// Payload for [`ResampleError::SourceChanged`].
1052///
1053/// A frame arrived whose shape is not the source spec this resampler
1054/// was built with — the mid-stream refusal.
1055///
1056/// The face never silently reconfigures: doing so would resample the
1057/// two halves of a stream on different terms and hand back a single
1058/// unbroken timeline built out of them. Build a new resampler for the
1059/// new source spec.
1060#[derive(thiserror::Error, Debug, Clone)]
1061#[error(
1062 "source format changed mid-stream: expected {expected_rate} Hz {expected_format:?}, \
1063 got {found_rate} Hz {found_format:?}"
1064)]
1065pub struct SourceChanged {
1066 expected_rate: u32,
1067 expected_format: SampleFormat,
1068 found_rate: u32,
1069 found_format: SampleFormat,
1070}
1071
1072impl SourceChanged {
1073 /// Constructs a `SourceChanged` payload.
1074 #[inline]
1075 pub const fn new(
1076 expected_rate: u32,
1077 expected_format: SampleFormat,
1078 found_rate: u32,
1079 found_format: SampleFormat,
1080 ) -> Self {
1081 Self {
1082 expected_rate,
1083 expected_format,
1084 found_rate,
1085 found_format,
1086 }
1087 }
1088 /// Rate the resampler was built for.
1089 #[inline]
1090 pub const fn expected_rate(&self) -> u32 {
1091 self.expected_rate
1092 }
1093 /// Sample format the resampler was built for.
1094 #[inline]
1095 pub const fn expected_format(&self) -> SampleFormat {
1096 self.expected_format
1097 }
1098 /// Rate the offending frame carried.
1099 #[inline]
1100 pub const fn found_rate(&self) -> u32 {
1101 self.found_rate
1102 }
1103 /// Sample format the offending frame carried.
1104 #[inline]
1105 pub const fn found_format(&self) -> SampleFormat {
1106 self.found_format
1107 }
1108}
1109
1110/// Payload for [`ResampleError::PlaneCount`].
1111///
1112/// A frame's planes do not hold what its header claims — too few
1113/// planes for the format, or a plane shorter than its sample count
1114/// requires.
1115#[derive(thiserror::Error, Debug, Clone)]
1116#[error("frame plane geometry mismatch: expected {expected}, found {found}")]
1117pub struct PlaneCount {
1118 expected: usize,
1119 found: usize,
1120}
1121
1122impl PlaneCount {
1123 /// Constructs a `PlaneCount` payload.
1124 #[inline]
1125 pub const fn new(expected: usize, found: usize) -> Self {
1126 Self { expected, found }
1127 }
1128 /// What the format and sample count require.
1129 #[inline]
1130 pub const fn expected(&self) -> usize {
1131 self.expected
1132 }
1133 /// What the frame carries.
1134 #[inline]
1135 pub const fn found(&self) -> usize {
1136 self.found
1137 }
1138}
1139
1140/// Payload for [`ResampleError::OutputTooLarge`].
1141///
1142/// The converted frame would be larger than
1143/// [`FrameLimits::max_frame_bytes`] allows.
1144///
1145/// Distinct from [`SampleCount`], which is about a count no `AVFrame`
1146/// can express at all. This one is about a frame FFmpeg would happily
1147/// allocate and the caller has said it does not want: the amplification
1148/// a hostile source rate buys — one second at a declared 1 Hz becomes
1149/// gigabytes at 192 kHz — lands here.
1150#[derive(thiserror::Error, Debug, Clone, Copy, PartialEq, Eq)]
1151#[error("a converted frame of {bytes} bytes exceeds the {limit}-byte ceiling")]
1152pub struct OutputTooLarge {
1153 bytes: usize,
1154 limit: usize,
1155}
1156
1157impl OutputTooLarge {
1158 /// Constructs an `OutputTooLarge` payload.
1159 #[cfg_attr(not(tarpaulin), inline(always))]
1160 pub const fn new(bytes: usize, limit: usize) -> Self {
1161 Self { bytes, limit }
1162 }
1163 /// The bytes the conversion would have produced.
1164 #[cfg_attr(not(tarpaulin), inline(always))]
1165 pub const fn bytes(&self) -> usize {
1166 self.bytes
1167 }
1168 /// The ceiling in force.
1169 #[cfg_attr(not(tarpaulin), inline(always))]
1170 pub const fn limit(&self) -> usize {
1171 self.limit
1172 }
1173}
1174
1175/// Payload for [`ResampleError::SampleCount`].
1176///
1177/// A sample count no frame can hold: one whose byte size overflows, or
1178/// one past the `c_int` `av_frame_get_buffer` takes.
1179#[derive(thiserror::Error, Debug, Clone)]
1180#[error("{requested} samples is not a frame size")]
1181pub struct SampleCount {
1182 requested: usize,
1183}
1184
1185impl SampleCount {
1186 /// Constructs a `SampleCount` payload.
1187 #[inline]
1188 pub const fn new(requested: usize) -> Self {
1189 Self { requested }
1190 }
1191 /// The count that was asked for.
1192 #[inline]
1193 pub const fn requested(&self) -> usize {
1194 self.requested
1195 }
1196}
1197
1198/// Payload for [`ResampleError::UnsupportedRate`].
1199///
1200/// One end of the conversion declares a sample rate `swr` cannot be
1201/// driven with — zero, or past `c_int`.
1202#[derive(thiserror::Error, Debug, Clone)]
1203#[error("the {end} rate {rate} is not a sample rate swr can use")]
1204pub struct UnsupportedRate {
1205 end: SpecEnd,
1206 rate: u32,
1207}
1208
1209impl UnsupportedRate {
1210 /// Constructs an `UnsupportedRate` payload.
1211 #[inline]
1212 pub const fn new(end: SpecEnd, rate: u32) -> Self {
1213 Self { end, rate }
1214 }
1215 /// Which end of the conversion.
1216 #[inline]
1217 pub const fn end(&self) -> SpecEnd {
1218 self.end
1219 }
1220 /// The rate that was declared.
1221 #[inline]
1222 pub const fn rate(&self) -> u32 {
1223 self.rate
1224 }
1225}
1226
1227/// Payload for [`ResampleError::UnsupportedFormat`].
1228///
1229/// One end of the conversion declares no sample format
1230/// (`AV_SAMPLE_FMT_NONE`) — the state a codec context is in before its
1231/// decoder opens.
1232#[derive(thiserror::Error, Debug, Clone)]
1233#[error("the {end} spec names no sample format")]
1234pub struct UnsupportedFormat {
1235 end: SpecEnd,
1236}
1237
1238impl UnsupportedFormat {
1239 /// Constructs an `UnsupportedFormat` payload.
1240 #[inline]
1241 pub const fn new(end: SpecEnd) -> Self {
1242 Self { end }
1243 }
1244 /// Which end of the conversion.
1245 #[inline]
1246 pub const fn end(&self) -> SpecEnd {
1247 self.end
1248 }
1249}
1250
1251/// Payload for [`ResampleError::UnsupportedLayout`].
1252///
1253/// One end of the conversion declares a channel layout this backend
1254/// will not carry.
1255///
1256/// Native and unspecified layouts are the two it does. A **custom** or
1257/// **ambisonic** `AVChannelLayout` owns a heap-allocated channel map,
1258/// and FFmpeg documents that such a layout must be copied with
1259/// `av_channel_layout_copy` rather than assigned — while
1260/// `ffmpeg_next::ChannelLayout` is a `Copy` wrapper with no destructor.
1261/// Every `AVFrame` this type stages or allocates receives the layout by
1262/// assignment, and `av_frame_free` runs `av_channel_layout_uninit` on
1263/// it: the first staged frame to be dropped would free a map the spec,
1264/// the decoder and every later frame still point at. Refusing at
1265/// construction is what keeps that use-after-free unreachable; a
1266/// resampler over those layouts is a separate design, not a silent
1267/// approximation.
1268#[derive(thiserror::Error, Debug, Clone)]
1269#[error("the {end} channel layout is not supported: order {order}, {channels} channels")]
1270pub struct UnsupportedLayout {
1271 end: SpecEnd,
1272 order: i32,
1273 channels: i32,
1274}
1275
1276impl UnsupportedLayout {
1277 /// Constructs an `UnsupportedLayout` payload.
1278 #[inline]
1279 pub const fn new(end: SpecEnd, order: i32, channels: i32) -> Self {
1280 Self {
1281 end,
1282 order,
1283 channels,
1284 }
1285 }
1286 /// Which end of the conversion.
1287 #[inline]
1288 pub const fn end(&self) -> SpecEnd {
1289 self.end
1290 }
1291 /// `AVChannelOrder` as the raw integer it is on the wire.
1292 #[inline]
1293 pub const fn order(&self) -> i32 {
1294 self.order
1295 }
1296 /// The channel count the layout declares.
1297 #[inline]
1298 pub const fn channels(&self) -> i32 {
1299 self.channels
1300 }
1301}
1302
1303/// Payload for [`ResampleError::TooManyPlanes`].
1304///
1305/// A planar spec with more channels than a decoded frame has plane
1306/// slots.
1307///
1308/// `mediadecode`'s `AudioFrame` carries a fixed eight planes
1309/// (`AV_NUM_DATA_POINTERS`); planar audio past that lives in
1310/// `AVFrame.extended_data[]`, which this crate does not plumb through.
1311/// As a **source** no valid frame could ever arrive; as a **target**
1312/// `swr` would produce one this crate cannot hand back — and it would
1313/// fail only after the input had been consumed, leaving a session that
1314/// cannot be retried. Both are refused at construction, where nothing
1315/// has happened yet.
1316#[derive(thiserror::Error, Debug, Clone)]
1317#[error("the {end} spec is planar with {channels} channels; a frame carries {limit} planes")]
1318pub struct TooManyPlanes {
1319 end: SpecEnd,
1320 channels: i32,
1321 limit: i32,
1322}
1323
1324impl TooManyPlanes {
1325 /// Constructs a `TooManyPlanes` payload.
1326 #[inline]
1327 pub const fn new(end: SpecEnd, channels: i32, limit: i32) -> Self {
1328 Self {
1329 end,
1330 channels,
1331 limit,
1332 }
1333 }
1334 /// Which end of the conversion.
1335 #[inline]
1336 pub const fn end(&self) -> SpecEnd {
1337 self.end
1338 }
1339 /// The channel count the layout declares.
1340 #[inline]
1341 pub const fn channels(&self) -> i32 {
1342 self.channels
1343 }
1344 /// Plane slots a frame has.
1345 #[inline]
1346 pub const fn limit(&self) -> i32 {
1347 self.limit
1348 }
1349}
1350
1351/// Payload for [`ResampleError::UnsupportedChannelCount`].
1352///
1353/// A spec declaring more channels than a frame's channel seat can
1354/// carry. `mediadecode`'s `AudioFrame` states its channel count in a
1355/// `u8`, so 255 is the ceiling in both directions.
1356///
1357/// This is the **packed** sibling of [`TooManyPlanes`], which only ever
1358/// caught planar specs: packed audio declares one plane whatever its
1359/// channel count, so a 256-channel packed spec sailed past that check
1360/// and had its count clipped on the way into the frame — a frame whose
1361/// bytes were computed from 256 channels while advertising 255. Refused
1362/// here, at the same choke point, for both ends.
1363#[derive(thiserror::Error, Debug, Clone)]
1364#[error("the {end} spec declares {channels} channels; a frame carries at most {limit}")]
1365pub struct UnsupportedChannelCount {
1366 end: SpecEnd,
1367 channels: i32,
1368 limit: i32,
1369}
1370
1371impl UnsupportedChannelCount {
1372 /// Constructs an `UnsupportedChannelCount` payload.
1373 #[inline]
1374 pub const fn new(end: SpecEnd, channels: i32, limit: i32) -> Self {
1375 Self {
1376 end,
1377 channels,
1378 limit,
1379 }
1380 }
1381 /// Which end of the conversion.
1382 #[inline]
1383 pub const fn end(&self) -> SpecEnd {
1384 self.end
1385 }
1386 /// The channel count the layout declares.
1387 #[inline]
1388 pub const fn channels(&self) -> i32 {
1389 self.channels
1390 }
1391 /// Channels a frame can state.
1392 #[inline]
1393 pub const fn limit(&self) -> i32 {
1394 self.limit
1395 }
1396}
1397
1398/// Payload for [`ResampleError::TimestampOutOfRange`].
1399///
1400/// A frame's timestamp does not land on the output timeline: it does
1401/// not survive the rescale as an `i64`, or it is `AV_NOPTS_VALUE`,
1402/// which is a sentinel rather than a time.
1403///
1404/// Raised before anything is staged, so a refused frame leaves the
1405/// resampler exactly as it was.
1406#[derive(thiserror::Error, Debug, Clone)]
1407#[error("the frame timestamp {pts} does not land on the output timeline")]
1408pub struct TimestampOutOfRange {
1409 pts: i64,
1410}
1411
1412impl TimestampOutOfRange {
1413 /// Constructs a `TimestampOutOfRange` payload.
1414 #[inline]
1415 pub const fn new(pts: i64) -> Self {
1416 Self { pts }
1417 }
1418 /// The timestamp the frame carried, in its own timebase.
1419 #[inline]
1420 pub const fn pts(&self) -> i64 {
1421 self.pts
1422 }
1423}
1424
1425/// Payload for [`ResampleError::ChannelDropped`].
1426///
1427/// The conversion between these two layouts would silently drop a
1428/// source channel: FFmpeg's own mixing matrix routes it to no output.
1429///
1430/// `swr` mixes the channel positions its rematrix table knows and
1431/// processes the rest of the input as though it were absent — a log
1432/// line at most. Measured against FFmpeg 9, packed 22.2 → mono loses
1433/// fifteen of twenty-four channels and `cube` → stereo loses two of
1434/// eight, so this is not a matter of channel count. Installing an
1435/// explicit mix matrix is how such a conversion would be accepted
1436/// deliberately; until this crate has a seat for one, the pair is
1437/// refused.
1438#[derive(thiserror::Error, Debug, Clone)]
1439#[error(
1440 "converting {source_channels} channels to {target_channels} would drop source channel \
1441 {channel}: FFmpeg's mixing matrix routes it to no output"
1442)]
1443pub struct ChannelDropped {
1444 source_channels: i32,
1445 target_channels: i32,
1446 channel: i32,
1447}
1448
1449impl ChannelDropped {
1450 /// Constructs a `ChannelDropped` payload.
1451 #[inline]
1452 pub const fn new(source_channels: i32, target_channels: i32, channel: i32) -> Self {
1453 Self {
1454 source_channels,
1455 target_channels,
1456 channel,
1457 }
1458 }
1459 /// Channels the source layout declares.
1460 #[inline]
1461 pub const fn source_channels(&self) -> i32 {
1462 self.source_channels
1463 }
1464 /// Channels the target layout declares.
1465 #[inline]
1466 pub const fn target_channels(&self) -> i32 {
1467 self.target_channels
1468 }
1469 /// The first source channel that reaches no output channel.
1470 #[inline]
1471 pub const fn channel(&self) -> i32 {
1472 self.channel
1473 }
1474}
1475
1476/// Payload for [`ResampleError::RematrixUnsupported`].
1477///
1478/// FFmpeg will not build a mixing matrix between these two layouts at
1479/// all.
1480#[derive(thiserror::Error, Debug, Clone)]
1481#[error("FFmpeg builds no mixing matrix from {source_channels} channels to {target_channels}")]
1482pub struct RematrixUnsupported {
1483 source_channels: i32,
1484 target_channels: i32,
1485}
1486
1487impl RematrixUnsupported {
1488 /// Constructs a `RematrixUnsupported` payload.
1489 #[inline]
1490 pub const fn new(source_channels: i32, target_channels: i32) -> Self {
1491 Self {
1492 source_channels,
1493 target_channels,
1494 }
1495 }
1496 /// Channels the source layout declares.
1497 #[inline]
1498 pub const fn source_channels(&self) -> i32 {
1499 self.source_channels
1500 }
1501 /// Channels the target layout declares.
1502 #[inline]
1503 pub const fn target_channels(&self) -> i32 {
1504 self.target_channels
1505 }
1506}
1507
1508/// Payload for [`ResampleError::TimestampOverflow`].
1509///
1510/// The output timeline would leave `i64`. Counted timestamps are exact
1511/// or they are nothing, so this is named rather than saturated.
1512#[derive(thiserror::Error, Debug, Clone)]
1513#[error("the output timeline overflows: {pts} + {samples} samples")]
1514pub struct TimestampOverflow {
1515 pts: i64,
1516 samples: i64,
1517}
1518
1519impl TimestampOverflow {
1520 /// Constructs a `TimestampOverflow` payload.
1521 #[inline]
1522 pub const fn new(pts: i64, samples: i64) -> Self {
1523 Self { pts, samples }
1524 }
1525 /// Where the timeline stood.
1526 #[inline]
1527 pub const fn pts(&self) -> i64 {
1528 self.pts
1529 }
1530 /// How many samples were produced.
1531 #[inline]
1532 pub const fn samples(&self) -> i64 {
1533 self.samples
1534 }
1535}
1536
1537/// Payload for [`ResampleError::OutputBuffer`].
1538///
1539/// A reference to one of the output frame's planes could not be taken.
1540///
1541/// Raised while preparing the conversion, never after it: that is the
1542/// point of preparing.
1543#[derive(thiserror::Error, Debug, Clone)]
1544#[error("the output frame's plane {plane} could not be referenced")]
1545pub struct OutputBuffer {
1546 plane: usize,
1547}
1548
1549impl OutputBuffer {
1550 /// Constructs an `OutputBuffer` payload.
1551 #[inline]
1552 pub const fn new(plane: usize) -> Self {
1553 Self { plane }
1554 }
1555 /// Which plane slot.
1556 #[inline]
1557 pub const fn plane(&self) -> usize {
1558 self.plane
1559 }
1560}
1561
1562/// Errors from [`FfmpegResampler`] — **faults and the two send-side
1563/// refusals** ([`Self::SourceChanged`], [`Self::AfterEof`]).
1564///
1565/// `Again` used to be here and meant two opposite things: pre-EOF
1566/// "nothing ready, send more" and post-EOF "the tail is exhausted".
1567/// A caller polling the seam could tell them apart only by remembering
1568/// whether it had itself called `send_eof`; one that did not spun
1569/// forever. Both are [`Received`] states now, and they are distinct.
1570///
1571/// **Open fault taxonomy, so it is `#[non_exhaustive]`.** New ways to
1572/// fail are discovered — a backend, a ceiling, a corruption a codec
1573/// learns to report — and a consumer that meets one it has never heard
1574/// of should take its generic-fault path. That is exactly what the
1575/// wildcard arm this attribute forces is for. The two status
1576/// vocabularies opposite it,
1577/// [`Sent`](mediadecode::Sent) and [`Received`](mediadecode::Received),
1578/// are exhaustive for the mirror-image reason: their arms are the
1579/// substrate's fixed state set, and there the wildcard would be dead
1580/// weight hiding a state a consumer forgot.
1581#[derive(thiserror::Error, Debug, Clone, IsVariant, Unwrap, TryUnwrap)]
1582#[unwrap(ref, ref_mut)]
1583#[try_unwrap(ref, ref_mut)]
1584#[non_exhaustive]
1585pub enum ResampleError {
1586 /// The conversion would produce a frame larger than the ceiling
1587 /// allows. Refused **before** the output frame is allocated.
1588 #[error(transparent)]
1589 OutputTooLarge(#[from] OutputTooLarge),
1590
1591 /// A frame arrived whose shape is not the source spec this resampler
1592 /// was built with — the mid-stream refusal.
1593 #[error(transparent)]
1594 SourceChanged(#[from] SourceChanged),
1595
1596 /// [`send_frame`](AudioResampler::send_frame) was called after
1597 /// [`send_eof`](AudioResampler::send_eof). Call
1598 /// [`flush`](AudioResampler::flush) first to reuse the resampler for
1599 /// another stream.
1600 #[error("send_frame after send_eof; flush() first to start another stream")]
1601 AfterEof,
1602
1603 /// A frame's planes do not hold what its header claims — too few
1604 /// planes for the format, or a plane shorter than its sample count
1605 /// requires.
1606 #[error(transparent)]
1607 PlaneCount(#[from] PlaneCount),
1608
1609 /// A sample count no frame can hold: one whose byte size overflows,
1610 /// or one past the `c_int` `av_frame_get_buffer` takes.
1611 #[error(transparent)]
1612 SampleCount(#[from] SampleCount),
1613
1614 /// One end of the conversion declares a sample rate `swr` cannot be
1615 /// driven with — zero, or past `c_int`.
1616 #[error(transparent)]
1617 UnsupportedRate(#[from] UnsupportedRate),
1618
1619 /// One end of the conversion declares no sample format
1620 /// (`AV_SAMPLE_FMT_NONE`) — the state a codec context is in before
1621 /// its decoder opens.
1622 #[error(transparent)]
1623 UnsupportedFormat(#[from] UnsupportedFormat),
1624
1625 /// One end of the conversion declares a channel layout this backend
1626 /// will not carry.
1627 #[error(transparent)]
1628 UnsupportedLayout(#[from] UnsupportedLayout),
1629
1630 /// A planar spec with more channels than a decoded frame has plane
1631 /// slots.
1632 #[error(transparent)]
1633 TooManyPlanes(#[from] TooManyPlanes),
1634
1635 /// A spec with more channels than a frame's channel seat can state.
1636 #[error(transparent)]
1637 UnsupportedChannelCount(#[from] UnsupportedChannelCount),
1638
1639 /// A frame's timestamp does not land on the output timeline.
1640 #[error(transparent)]
1641 TimestampOutOfRange(#[from] TimestampOutOfRange),
1642
1643 /// The conversion between these two layouts would silently drop a
1644 /// source channel.
1645 #[error(transparent)]
1646 ChannelDropped(#[from] ChannelDropped),
1647
1648 /// FFmpeg will not build a mixing matrix between these two layouts at
1649 /// all.
1650 #[error(transparent)]
1651 RematrixUnsupported(#[from] RematrixUnsupported),
1652
1653 /// The output timeline would leave `i64`. Counted timestamps are
1654 /// exact or they are nothing, so this is named rather than saturated.
1655 #[error(transparent)]
1656 TimestampOverflow(#[from] TimestampOverflow),
1657
1658 /// The wrapped `swresample` call reported an error.
1659 #[error(transparent)]
1660 Resample(#[from] Error),
1661
1662 /// A reference to one of the output frame's planes could not be
1663 /// taken.
1664 #[error(transparent)]
1665 OutputBuffer(#[from] OutputBuffer),
1666
1667 /// The queue of converted frames could not be grown to hold one more.
1668 #[error("out of memory reserving room for a converted frame")]
1669 QueueAlloc,
1670}
1671
1672/// Which end of a conversion a refusal is about.
1673#[derive(Copy, Clone, Debug, PartialEq, Eq, IsVariant)]
1674pub enum SpecEnd {
1675 /// The spec frames must arrive in.
1676 Source,
1677 /// The spec frames leave in.
1678 Target,
1679}
1680
1681impl core::fmt::Display for SpecEnd {
1682 fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
1683 f.write_str(match self {
1684 Self::Source => "source",
1685 Self::Target => "target",
1686 })
1687 }
1688}
1689
1690/// Plane slots a `mediadecode::frame::AudioFrame` has — the fixed array
1691/// matching `AV_NUM_DATA_POINTERS`. Planar audio past this many
1692/// channels lives in `AVFrame.extended_data[]` / `extended_buf[]`,
1693/// which this crate does not plumb through: `convert` refuses such a
1694/// frame and `AudioFrame::new` will not build one.
1695const MAX_AUDIO_PLANES: i32 = 8;
1696
1697/// Channels a `mediadecode::frame::AudioFrame` can state — its channel
1698/// seat is a `u8`. A spec past this is refused rather than clipped.
1699const MAX_FRAME_CHANNELS: i32 = u8::MAX as i32;
1700
1701/// Refuses a spec `swr` cannot be driven with, or whose channel layout
1702/// cannot be carried by value — see [`ResampleError::UnsupportedLayout`]
1703/// for that one, which is the whole reason this check exists at the
1704/// choke point rather than in the `const` constructor.
1705fn check_spec(spec: &ResampleSpec, end: SpecEnd) -> Result<(), ResampleError> {
1706 if spec.rate == 0 || spec.rate > i32::MAX as u32 {
1707 return Err(ResampleError::UnsupportedRate(UnsupportedRate::new(
1708 end, spec.rate,
1709 )));
1710 }
1711 if spec.format == Sample::None {
1712 return Err(ResampleError::UnsupportedFormat(UnsupportedFormat::new(
1713 end,
1714 )));
1715 }
1716 let order = layout_order(&spec.layout);
1717 let channels = spec.layout.channels();
1718 let carried = order == AVChannelOrder::AV_CHANNEL_ORDER_NATIVE as i32
1719 || order == AVChannelOrder::AV_CHANNEL_ORDER_UNSPEC as i32;
1720 if !carried || channels <= 0 {
1721 return Err(ResampleError::UnsupportedLayout(UnsupportedLayout::new(
1722 end, order, channels,
1723 )));
1724 }
1725 // A planar spec with more channels than the frame model has plane
1726 // slots is a resampler that cannot work in either direction, and
1727 // saying so here is the difference between a refusal at construction
1728 // and a refusal on every frame — the target one arriving *after*
1729 // `swr` has already consumed the input, which is not a state a caller
1730 // can retry from.
1731 if spec.format.is_planar() && channels > MAX_AUDIO_PLANES {
1732 return Err(ResampleError::TooManyPlanes(TooManyPlanes::new(
1733 end,
1734 channels,
1735 MAX_AUDIO_PLANES,
1736 )));
1737 }
1738 // And the packed sibling, which the plane check above cannot see: a
1739 // packed spec declares one plane at any channel count, so it reached
1740 // the frame with its count clipped to 255 instead of refused. That is
1741 // the same silent truncation the decode path was carrying, on the
1742 // other audio road — closed here, at the same choke point, so that
1743 // every channel count downstream is exact by construction.
1744 if channels > MAX_FRAME_CHANNELS {
1745 return Err(ResampleError::UnsupportedChannelCount(
1746 UnsupportedChannelCount::new(end, channels, MAX_FRAME_CHANNELS),
1747 ));
1748 }
1749 Ok(())
1750}
1751
1752/// A layout's `AVChannelOrder` as the integer it is on the wire.
1753///
1754/// Read raw rather than matched as an `AVChannelOrder`, the discipline
1755/// this crate keeps everywhere it touches a bindgen enum: a value
1756/// outside this build's discriminant set would be undefined behaviour
1757/// the moment it existed as one.
1758fn layout_order(layout: &ChannelLayout) -> i32 {
1759 // SAFETY: `layout` is a live `ChannelLayout` for the duration of this
1760 // call; `addr_of!` reaches its `order` field without forming a
1761 // reference to the enum.
1762 unsafe { read_unaligned(addr_of!(layout.0.order).cast::<i32>()) }
1763}
1764
1765/// FFmpeg's `SWR_CH_MAX`: the square its own matrix builder writes,
1766/// whatever the two layouts' channel counts are.
1767///
1768/// Not a convenience. `swr_build_matrix2` copies its internal
1769/// `[SWR_CH_MAX][SWR_CH_MAX]` block out at the caller's stride, so a
1770/// buffer sized to the actual channel counts is written far past its
1771/// end — measured, and the measurement is a killed process.
1772const SWR_CH_MAX: usize = 64;
1773
1774/// Refuses an **effective pair** whose rematrixing would silently drop
1775/// input channels.
1776///
1777/// Takes the layouts `swr` is configured with, not the ones the caller
1778/// declared. The two differ exactly where it matters: an unspecified
1779/// layout is resolved to FFmpeg's default for its channel count before
1780/// the context is opened, and twenty-four unspecified channels resolve
1781/// to 22.2 — so the declared pair says "unspecified, nothing to
1782/// rematrix" while the conversion that runs is the lossy one.
1783///
1784/// This is the second half of the crate's two-layout bookkeeping, and
1785/// the halves answer different questions. The **declared** layout is
1786/// what decoded frames carry (a WAV without a channel mask hands out
1787/// unspecified frames forever) and stays the yardstick for the
1788/// mid-stream refusal: *is this frame the stream I was built for?* The
1789/// **effective** layout is what `swr` and every staged `AVFrame` use,
1790/// and it is the one judged here: *what will `swr` actually do?*
1791///
1792/// Each end can be perfectly valid on its own and the conversion
1793/// between them still lose whole channels: `swr` mixes only the channel
1794/// positions its rematrix table knows, and quietly processes the rest
1795/// of the input as though it were not there. Measured against the
1796/// linked FFmpeg 9 with a tone isolated in each source channel: packed
1797/// 22.2 → mono drops fifteen of twenty-four (`swr` says as much in a
1798/// log line and converts anyway), `cube` → stereo drops two of *eight*
1799/// — so a channel-count threshold is both too strict and too loose to
1800/// be the rule.
1801///
1802/// The rule is asked of FFmpeg instead: build the mixing matrix its own
1803/// builder would use, and refuse when any input channel reaches no
1804/// output at all. `lfe_mix_level` is deliberately non-zero, so the
1805/// question is "can this channel reach the output" rather than "does
1806/// FFmpeg's default downmix policy include it" — the default leaves LFE
1807/// out of a downmix on purpose, and refusing an everyday 5.1 → stereo
1808/// over that would be absurd. The predicate matched the tone sweep
1809/// exactly on every pair measured.
1810///
1811/// A pair `swr` cannot matrix at all is refused too. Accepting these
1812/// deliberately is a *mix matrix* seat on the spec — a real design, not
1813/// something to mint in passing; until it exists, refusal is the honest
1814/// answer.
1815fn check_pair(source: &ChannelLayout, target: &ChannelLayout) -> Result<(), ResampleError> {
1816 let native = AVChannelOrder::AV_CHANNEL_ORDER_NATIVE as i32;
1817 // A layout still unspecified *after* resolution — a channel count
1818 // FFmpeg has no default for — is mapped positionally by `swr` with no
1819 // rematrixing at all, and identical layouts need no matrix: neither
1820 // can drop a channel, and neither is what the builder describes.
1821 if layout_order(source) != native || layout_order(target) != native || source == target {
1822 return Ok(());
1823 }
1824 let source_channels = source.channels();
1825 let target_channels = target.channels();
1826
1827 let mut matrix = vec![0f64; SWR_CH_MAX * SWR_CH_MAX];
1828 // SAFETY: both layouts are live for the call; `matrix` is the full
1829 // `SWR_CH_MAX` square the builder writes, passed with the matching
1830 // stride; the encoding is a compile-time constant of this build; and
1831 // a null log context is documented as allowed.
1832 let rc = unsafe {
1833 swr_build_matrix2(
1834 &source.0,
1835 &target.0,
1836 core::f64::consts::FRAC_1_SQRT_2,
1837 core::f64::consts::FRAC_1_SQRT_2,
1838 1.0,
1839 1.0,
1840 1.0,
1841 matrix.as_mut_ptr(),
1842 SWR_CH_MAX as isize,
1843 AVMatrixEncoding::AV_MATRIX_ENCODING_NONE,
1844 core::ptr::null_mut(),
1845 )
1846 };
1847 if rc < 0 {
1848 return Err(ResampleError::RematrixUnsupported(
1849 RematrixUnsupported::new(source_channels, target_channels),
1850 ));
1851 }
1852 for channel in 0..source_channels.min(SWR_CH_MAX as i32) {
1853 let index = channel as usize;
1854 if (0..target_channels.min(SWR_CH_MAX as i32) as usize)
1855 .all(|out| matrix[index + SWR_CH_MAX * out] == 0.0)
1856 {
1857 return Err(ResampleError::ChannelDropped(ChannelDropped::new(
1858 source_channels,
1859 target_channels,
1860 channel,
1861 )));
1862 }
1863 }
1864 Ok(())
1865}
1866
1867/// Opens a `swr` context for the two specs. Shared by
1868/// [`FfmpegResampler::new`] and the rebuild
1869/// [`AudioResampler::flush`] performs.
1870fn open_context(
1871 source: &ResampleSpec,
1872 target: &ResampleSpec,
1873 staged_source_layout: ChannelLayout,
1874 staged_target_layout: ChannelLayout,
1875) -> Result<resampling::Context, ResampleError> {
1876 resampling::Context::get(
1877 source.format,
1878 staged_source_layout,
1879 source.rate,
1880 target.format,
1881 staged_target_layout,
1882 target.rate,
1883 )
1884 .map_err(|e| ResampleError::Resample(Error::Ffmpeg(e)))
1885}
1886
1887/// Allocates an audio `AVFrame`, checking every step the dependency's
1888/// own `frame::Audio::new` does not.
1889///
1890/// `ffmpeg_next`'s constructor dereferences `av_frame_alloc`'s result
1891/// without a null check and discards `av_frame_get_buffer`'s return
1892/// value, so an allocation failure there yields a frame whose planes
1893/// are not backed — which is then handed to FFI. Both are checked here;
1894/// on failure the caller gets a named error and no frame at all. The
1895/// null check is the crate's existing one
1896/// ([`crate::frame::alloc_av_audio_frame`], which the decoders already
1897/// allocate through), so there is one answer to `av_frame_alloc`
1898/// returning null rather than two.
1899fn new_audio_frame(
1900 format: Sample,
1901 samples: usize,
1902 rate: u32,
1903 layout: ChannelLayout,
1904) -> Result<frame::Audio, ResampleError> {
1905 if samples == 0 || samples > i32::MAX as usize {
1906 return Err(ResampleError::SampleCount(SampleCount::new(samples)));
1907 }
1908 let mut out = crate::frame::alloc_av_audio_frame()?;
1909 out.set_format(format);
1910 out.set_samples(samples);
1911 // The layout is assigned by value, which is sound only because
1912 // `check_spec` refused every layout that owns a heap channel map.
1913 out.set_channel_layout(layout);
1914 out.set_rate(rate);
1915 // SAFETY: `out` is a live `AVFrame` whose format, sample count and
1916 // layout were just set; `av_frame_get_buffer` allocates its planes
1917 // and reports failure in its return value, which is checked.
1918 let rc = unsafe { av_frame_get_buffer(out.as_mut_ptr(), 0) };
1919 if rc < 0 {
1920 return Err(ResampleError::Resample(Error::Ffmpeg(
1921 ffmpeg_next::Error::from(rc),
1922 )));
1923 }
1924 Ok(out)
1925}
1926
1927/// Bytes one plane holds for `samples` samples of `format`, or `None`
1928/// when that product does not fit a `usize`. Packed formats keep every
1929/// channel in the single plane; planar formats give each channel its
1930/// own.
1931fn plane_bytes(format: Sample, samples: usize, channels: i32) -> Option<usize> {
1932 // Refused, not floored. `check_spec` already proves the count is in
1933 // `1..=MAX_FRAME_CHANNELS` before any resampler exists, so this is a
1934 // restatement of an invariant rather than a live branch — but it is
1935 // stated as a refusal because the alternative was a substituted `1`,
1936 // which invents a channel the caller never declared and makes the
1937 // byte product disagree with the frame it sizes.
1938 let channels = usize::try_from(channels).ok().filter(|count| *count > 0)?;
1939 let bytes = samples.checked_mul(format.bytes())?;
1940 if format.is_planar() {
1941 Some(bytes)
1942 } else {
1943 bytes.checked_mul(channels)
1944 }
1945}
1946
1947/// Builds a native-order [`ChannelLayout`] from a channel bitmask,
1948/// without ever forming an `AVChannelLayout` out of foreign memory:
1949/// the struct starts zeroed (`AV_CHANNEL_ORDER_UNSPEC` is `0`, a valid
1950/// discriminant) and FFmpeg fills it.
1951fn layout_from_mask(mask: u64) -> ChannelLayout {
1952 // SAFETY: a zeroed `AVChannelLayout` is a valid value — its `order`
1953 // field reads as `AV_CHANNEL_ORDER_UNSPEC`, the zero discriminant —
1954 // and `av_channel_layout_from_mask` overwrites it wholesale.
1955 unsafe {
1956 let mut layout = std::mem::zeroed();
1957 if av_channel_layout_from_mask(&mut layout, mask) < 0 {
1958 return ChannelLayout::default(mask.count_ones() as i32);
1959 }
1960 ChannelLayout(layout)
1961 }
1962}
1963
1964/// The layout `swr` will actually be configured with.
1965///
1966/// `swr_init` replaces an unspecified input or output layout with
1967/// FFmpeg's default for that channel count, and from then on compares
1968/// every frame handed to it against *that* layout — a staged frame
1969/// still carrying the unspecified one is refused with
1970/// `AVERROR_INPUT_CHANGED`. Applying the same rule here, once, keeps
1971/// the frames this type builds in step with the context it built.
1972///
1973/// The declared layout is kept separately and is what the mid-stream
1974/// check compares against, because it is what decoded frames really
1975/// carry: a WAV without a channel mask hands out unspecified frames
1976/// forever, whatever `swr` decided internally.
1977fn initialized_layout(layout: ChannelLayout) -> ChannelLayout {
1978 if layout.is_empty() {
1979 ChannelLayout::default(layout.channels())
1980 } else {
1981 layout
1982 }
1983}
1984
1985/// Reads an `AVChannelLayout` out of FFmpeg memory into a layout this
1986/// spec can own, or `None` for one it does not represent.
1987///
1988/// The `order` field is read as the integer it is on the wire: an
1989/// out-of-range value would be undefined behaviour the instant it
1990/// existed as an `AVChannelOrder`, which is the hazard this crate
1991/// keeps out everywhere it touches a bindgen enum.
1992///
1993/// A **custom** or **ambisonic** layout returns `None`. Both keep a
1994/// heap-allocated channel map inside the layout, and `ChannelLayout` is
1995/// a plain `Copy` wrapper with no destructor: owning one here would
1996/// either alias a map the decoder still frees or leak the copy. A
1997/// resampler over one of those layouts is a separate design, not a
1998/// silent approximation.
1999///
2000/// # Safety
2001///
2002/// `ptr` must be a live `*const AVChannelLayout` for the duration of
2003/// this call.
2004unsafe fn layout_from_raw(ptr: *const ffmpeg_next::ffi::AVChannelLayout) -> Option<ChannelLayout> {
2005 let order = unsafe { read_unaligned(addr_of!((*ptr).order).cast::<i32>()) };
2006 let channels = unsafe { (*ptr).nb_channels };
2007 // The union's `mask` arm is only read for the order that names it —
2008 // for `CUSTOM` those eight bytes are a pointer. `layout_from_parts`
2009 // refuses that order anyway, so a zero is the honest thing to hand
2010 // it and reading the pointer would be the dishonest one.
2011 let mask = if order == AVChannelOrder::AV_CHANNEL_ORDER_NATIVE as i32 {
2012 // SAFETY: `u.mask` is the union's variant for NATIVE, and the
2013 // order was checked against our own constant before the read.
2014 unsafe { (*ptr).u.mask }
2015 } else {
2016 0
2017 };
2018 layout_from_parts(order, channels, mask)
2019}
2020
2021/// The layout roster, over parts rather than a pointer.
2022///
2023/// One function so the two roads into a source spec —
2024/// [`ResampleSpec::from_parameters`] over a live `AVCodecParameters`
2025/// and [`ResampleSpec::from_ticket`] over an owned mirror — cannot
2026/// drift into admitting different layouts from the same file.
2027fn layout_from_parts(order: i32, channels: i32, mask: u64) -> Option<ChannelLayout> {
2028 if channels <= 0 {
2029 return None;
2030 }
2031 if order == AVChannelOrder::AV_CHANNEL_ORDER_NATIVE as i32 {
2032 if mask != 0 {
2033 return Some(layout_from_mask(mask));
2034 }
2035 // Native in name with no channels named: unspecified in substance.
2036 return Some(ResampleSpec::unspecified_layout(channels));
2037 }
2038 if order == AVChannelOrder::AV_CHANNEL_ORDER_UNSPEC as i32 {
2039 return Some(ResampleSpec::unspecified_layout(channels));
2040 }
2041 None
2042}
2043
2044/// Compile-time assurance that `SampleFormat`'s round trip through
2045/// FFmpeg's vocabulary is the identity on the closed set. Both
2046/// directions are hand-written tables, and a table that disagreed with
2047/// its inverse would silently mislabel every sample.
2048const _: () = {
2049 assert!(
2050 SampleFormat::from_raw(AVSampleFormat::AV_SAMPLE_FMT_NONE as i32)
2051 .to_ffmpeg()
2052 .is_none()
2053 );
2054};
2055
2056#[cfg(test)]
2057mod tests {
2058 use super::*;
2059
2060 use mediadecode::resampler::AudioResampler;
2061
2062 // These exercise the conversion arithmetic — rates, layouts, the
2063 // counted timeline, the byte ceiling — which is lane-independent, so
2064 // they run on the owned lane exactly as they did before the second
2065 // lane existed. The view lane's own road through this type (reserve
2066 // before `swr`, commit after) is proved in `tests/view_carriers.rs`,
2067 // where a produced plane can be shown to point into the output
2068 // frame's buffer.
2069 use crate::{FfmpegBytes, FfmpegOwnedResampler as FfmpegResampler};
2070
2071 type Frame = super::Frame<crate::Owned>;
2072
2073 /// A 48 kHz packed-s16 stereo frame of silence, with the plane its
2074 /// header claims.
2075 fn stereo_frame(samples: u32) -> Frame {
2076 let plane = FfmpegBytes::copy_from_slice(&vec![0u8; samples as usize * 2 * 2]);
2077 let planes = std::array::from_fn(|index| {
2078 Plane::new(
2079 if index == 0 {
2080 plane.clone()
2081 } else {
2082 FfmpegBytes::empty()
2083 },
2084 0,
2085 )
2086 });
2087 AudioFrame::new(
2088 48_000,
2089 samples,
2090 2,
2091 SampleFormat::S16,
2092 crate::channel_layout::channel_layout_description_from_ffmpeg(&ChannelLayout::STEREO),
2093 planes,
2094 1,
2095 AudioFrameExtra::default(),
2096 )
2097 .with_pts(Some(Timestamp::new(
2098 0,
2099 Timebase::new(1, std::num::NonZeroI32::new(48_000).expect("a real rate")),
2100 )))
2101 }
2102
2103 /// Takes every frame that is ready *right now* and stops — the
2104 /// pre-EOF half of a drain, written as the exhaustive match the face
2105 /// now requires. A `.is_ok()` loop here would never end: "needs
2106 /// input" is a success.
2107 fn drain_ready(resampler: &mut FfmpegResampler, dst: &mut Frame) {
2108 loop {
2109 match resampler.receive_frame(dst).expect("a fault-free drain") {
2110 Received::Frame => {}
2111 Received::NeedsInput | Received::Ended => return,
2112 }
2113 }
2114 }
2115
2116 fn stereo_to_mono() -> FfmpegResampler {
2117 FfmpegResampler::new(
2118 ResampleSpec::new(
2119 48_000,
2120 Sample::I16(ffmpeg_next::format::sample::Type::Packed),
2121 ChannelLayout::STEREO,
2122 ),
2123 ResampleSpec::new(
2124 16_000,
2125 Sample::I16(ffmpeg_next::format::sample::Type::Packed),
2126 ChannelLayout::MONO,
2127 ),
2128 FrameLimits::default(),
2129 )
2130 .expect("open resampler")
2131 }
2132
2133 #[test]
2134 fn resample_error_carries_the_derived_accessor_face() {
2135 // `IsVariant` / `Unwrap` / `TryUnwrap` — one arm per derive family,
2136 // mirroring the mediadecode-side proof for this crate's own
2137 // newly-wired `derive_more` dependency.
2138 let err = ResampleError::OutputBuffer(OutputBuffer::new(2));
2139 assert!(err.is_output_buffer());
2140 assert!(!err.is_queue_alloc());
2141 assert_eq!(err.unwrap_output_buffer_ref().plane(), 2);
2142 assert!(err.try_unwrap_queue_alloc().is_err());
2143 }
2144
2145 #[test]
2146 fn an_allocation_fault_while_sending_leaves_the_session_untouched() {
2147 // The class this design exists to end: a failure on the far side of
2148 // `swr_convert_frame` leaves a session no caller can act on —
2149 // retrying feeds the same samples twice, continuing loses them, and
2150 // the delay line has moved either way. Every allocation the
2151 // conversion needs is taken before `swr` runs, so an allocator that
2152 // refuses everything can only produce an error that cost nothing.
2153 crate::fault_subprocess::in_subprocess(
2154 "resampler::tests::an_allocation_fault_while_sending_leaves_the_session_untouched",
2155 || {
2156 let mut resampler = stereo_to_mono();
2157 let frame = stereo_frame(4_800);
2158 let mut dst = crate::boundary::empty_owned_audio_frame();
2159 crate::accepted(resampler.send_frame(&frame), "a first frame");
2160 drain_ready(&mut resampler, &mut dst);
2161 let delay = resampler.delay();
2162 assert!(delay > 0, "the filter has to be holding something");
2163
2164 crate::fault_subprocess::cap_ffmpeg_allocations(1);
2165 let refused = resampler.send_frame(&frame);
2166 crate::fault_subprocess::uncap_ffmpeg_allocations();
2167
2168 assert!(
2169 refused.is_err(),
2170 "an allocator that refuses everything must not look like success",
2171 );
2172 assert_eq!(
2173 resampler.delay(),
2174 delay,
2175 "the frame went into the filter anyway",
2176 );
2177 assert!(
2178 matches!(resampler.receive_frame(&mut dst), Ok(Received::NeedsInput)),
2179 "a failed send left output ready",
2180 );
2181
2182 // And the session is still a session: the same frame converts.
2183 crate::accepted(resampler.send_frame(&frame), "the failure cost nothing");
2184 assert!(matches!(
2185 resampler.receive_frame(&mut dst),
2186 Ok(Received::Frame)
2187 ));
2188 },
2189 );
2190 }
2191
2192 #[test]
2193 fn an_allocation_fault_while_draining_keeps_the_tail() {
2194 // The same property one call along, where the samples at risk are
2195 // the ones already inside the filter: a drain that fails must leave
2196 // the tail where it was, not turn it into an error.
2197 crate::fault_subprocess::in_subprocess(
2198 "resampler::tests::an_allocation_fault_while_draining_keeps_the_tail",
2199 || {
2200 let mut resampler = stereo_to_mono();
2201 let frame = stereo_frame(4_800);
2202 let mut dst = crate::boundary::empty_owned_audio_frame();
2203 for _ in 0..3 {
2204 crate::accepted(resampler.send_frame(&frame), "send_frame");
2205 drain_ready(&mut resampler, &mut dst);
2206 }
2207 crate::accepted(resampler.send_eof(), "eof");
2208 let tail = resampler.delay();
2209 assert!(tail > 0, "there has to be a tail to lose");
2210
2211 crate::fault_subprocess::cap_ffmpeg_allocations(1);
2212 let refused = resampler.receive_frame(&mut dst);
2213 crate::fault_subprocess::uncap_ffmpeg_allocations();
2214
2215 let refused = refused.expect_err("the drain cannot have succeeded");
2216 // No arm of this enum can say "send me more input" any more —
2217 // that answer left the error type — so an allocation failure has
2218 // nowhere to be mistaken for one.
2219 assert!(
2220 matches!(
2221 refused,
2222 ResampleError::Resample(_) | ResampleError::OutputBuffer(_)
2223 ),
2224 "an allocation failure surfaced as something else: {refused:?}",
2225 );
2226 assert_eq!(
2227 resampler.delay(),
2228 tail,
2229 "the tail was consumed by a drain that failed",
2230 );
2231
2232 // And it is still drainable, which is the whole point.
2233 assert_eq!(
2234 resampler
2235 .receive_frame(&mut dst)
2236 .expect("the tail survived the failure"),
2237 Received::Frame,
2238 );
2239 },
2240 );
2241 }
2242
2243 #[test]
2244 fn the_sample_format_table_round_trips() {
2245 for format in [
2246 SampleFormat::U8,
2247 SampleFormat::S16,
2248 SampleFormat::S32,
2249 SampleFormat::S64,
2250 SampleFormat::FLT,
2251 SampleFormat::DBL,
2252 SampleFormat::U8P,
2253 SampleFormat::S16P,
2254 SampleFormat::S32P,
2255 SampleFormat::S64P,
2256 SampleFormat::FLTP,
2257 SampleFormat::DBLP,
2258 ] {
2259 let ffmpeg = format.to_ffmpeg().expect("a named format");
2260 assert_eq!(
2261 SampleFormat::from_ffmpeg(ffmpeg),
2262 format,
2263 "{format:?} does not survive the round trip",
2264 );
2265 assert_eq!(ffmpeg.is_planar(), format.is_planar());
2266 }
2267 assert!(SampleFormat::NONE.to_ffmpeg().is_none());
2268 assert!(SampleFormat::from_raw(9999).to_ffmpeg().is_none());
2269 }
2270
2271 #[test]
2272 fn a_mask_rebuilds_the_layout_it_names() {
2273 let stereo = layout_from_mask(ChannelLayout::STEREO.bits());
2274 assert_eq!(stereo.channels(), 2);
2275 assert_eq!(stereo.bits(), ChannelLayout::STEREO.bits());
2276
2277 let five_one = layout_from_mask(ChannelLayout::_5POINT1.bits());
2278 assert_eq!(five_one.channels(), 6);
2279 assert_eq!(
2280 five_one.bits(),
2281 ChannelLayout::_5POINT1.bits(),
2282 "the side-vs-back distinction is exactly what a default layout would lose",
2283 );
2284 }
2285
2286 #[test]
2287 fn plane_geometry_follows_packed_versus_planar() {
2288 use ffmpeg_next::format::sample::Type;
2289 // Packed: one plane holding every channel.
2290 assert_eq!(
2291 plane_bytes(Sample::I16(Type::Packed), 1024, 2),
2292 Some(1024 * 2 * 2)
2293 );
2294 // Planar: one plane per channel, so the count does not multiply in.
2295 assert_eq!(
2296 plane_bytes(Sample::I16(Type::Planar), 1024, 2),
2297 Some(1024 * 2)
2298 );
2299 assert_eq!(
2300 plane_bytes(Sample::F32(Type::Planar), 1024, 6),
2301 Some(1024 * 4)
2302 );
2303 // A sample count whose byte size does not fit is not a size. This
2304 // is the arithmetic that used to run before the allocation it
2305 // feeds, and it wrapped.
2306 assert_eq!(
2307 plane_bytes(Sample::F32(Type::Packed), usize::MAX / 2, 8),
2308 None,
2309 "an overflowing plane size is refused, not wrapped",
2310 );
2311 }
2312
2313 #[test]
2314 fn the_target_timebase_is_one_tick_per_output_sample() {
2315 let spec = ResampleSpec::new(
2316 16_000,
2317 Sample::I16(ffmpeg_next::format::sample::Type::Packed),
2318 ChannelLayout::MONO,
2319 );
2320 let tb = spec.timebase();
2321 assert_eq!((tb.num(), tb.den().get()), (1, 16_000));
2322 }
2323}