Hi
On Mon, Feb 16, 2026 at 10:32 AM Mark Cave-Ayland
<[email protected]> wrote:
>
> On 14/02/2026 08:20, Marc-André Lureau wrote:
>
> > Hi
> >
> > On Fri, Feb 13, 2026 at 4:28 PM Mark Cave-Ayland
> > <[email protected]> wrote:
> >>
> >> On 27/01/2026 18:24, [email protected] wrote:
> >>
> >>> From: Marc-André Lureau <[email protected]>
> >>>
> >>> The audio_pcm_info structure stored three fields (bits, is_signed,
> >>> is_float) that were always derived from the AudioFormat enum. This
> >>> redundancy meant the same information was represented twice, with no
> >>> type-level guarantee that they stayed in sync.
> >>>
> >>> Replace these fields with a single AudioFormat field, and add helper
> >>> functions to extract the derived properties when needed:
> >>> - audio_format_bits()
> >>> - audio_format_is_signed()
> >>> - audio_format_is_float()
> >>>
> >>> This improves type safety by making AudioFormat the single source of
> >>> truth, eliminating the possibility of inconsistent state between the
> >>> format enum and its derived boolean/integer representations.
> >>>
> >>> Signed-off-by: Marc-André Lureau <[email protected]>
> >>> ---
> >>> audio/audio_int.h | 4 +-
> >>> audio/audio_template.h | 12 +--
> >>> include/qemu/audio.h | 49 +++++++++
> >>> audio/audio-mixeng-be.c | 218 ++++++++++++----------------------------
> >>> audio/dbusaudio.c | 12 +--
> >>> audio/coreaudio.m | 2 +-
> >>> 6 files changed, 126 insertions(+), 171 deletions(-)
> >>>
> >>> diff --git a/audio/audio_int.h b/audio/audio_int.h
> >>> index 5334c4baad2..dd5f2220d75 100644
> >>> --- a/audio/audio_int.h
> >>> +++ b/audio/audio_int.h
> >>> @@ -45,9 +45,7 @@ struct audio_callback {
> >>> };
> >>>
> >>> struct audio_pcm_info {
> >>> - int bits;
> >>> - bool is_signed;
> >>> - bool is_float;
> >>> + AudioFormat af;
> >>> int freq;
> >>> int nchannels;
> >>> int bytes_per_frame;
> >>> diff --git a/audio/audio_template.h b/audio/audio_template.h
> >>> index 08d60422589..3da91a4782c 100644
> >>> --- a/audio/audio_template.h
> >>> +++ b/audio/audio_template.h
> >>> @@ -173,7 +173,7 @@ static int glue (audio_pcm_sw_init_, TYPE) (
> >>> sw->empty = true;
> >>> #endif
> >>>
> >>> - if (sw->info.is_float) {
> >>> + if (audio_format_is_float(hw->info.af)) {
> >>> #ifdef DAC
> >>> sw->conv = mixeng_conv_float[sw->info.nchannels == 2]
> >>> [sw->info.swap_endianness];
> >>> @@ -188,9 +188,9 @@ static int glue (audio_pcm_sw_init_, TYPE) (
> >>> sw->clip = mixeng_clip
> >>> #endif
> >>> [sw->info.nchannels == 2]
> >>> - [sw->info.is_signed]
> >>> + [audio_format_is_signed(hw->info.af)]
> >>> [sw->info.swap_endianness]
> >>> - [audio_bits_to_index(sw->info.bits)];
> >>> + [audio_format_to_index(hw->info.af)];
> >>> }
> >>>
> >>> sw->name = g_strdup (name);
> >>> @@ -300,7 +300,7 @@ static HW *glue(audio_pcm_hw_add_new_,
> >>> TYPE)(AudioMixengBackend *s,
> >>> goto err1;
> >>> }
> >>>
> >>> - if (hw->info.is_float) {
> >>> + if (audio_format_is_float(hw->info.af)) {
> >>> #ifdef DAC
> >>> hw->clip = mixeng_clip_float[hw->info.nchannels == 2]
> >>> [hw->info.swap_endianness];
> >>> @@ -315,9 +315,9 @@ static HW *glue(audio_pcm_hw_add_new_,
> >>> TYPE)(AudioMixengBackend *s,
> >>> hw->conv = mixeng_conv
> >>> #endif
> >>> [hw->info.nchannels == 2]
> >>> - [hw->info.is_signed]
> >>> + [audio_format_is_signed(hw->info.af)]
> >>> [hw->info.swap_endianness]
> >>> - [audio_bits_to_index(hw->info.bits)];
> >>> + [audio_format_to_index(hw->info.af)];
> >>> }
> >>>
> >>> glue(audio_pcm_hw_alloc_resources_, TYPE)(hw);
> >>> diff --git a/include/qemu/audio.h b/include/qemu/audio.h
> >>> index b6b6ee9b560..a2fbc286eb1 100644
> >>> --- a/include/qemu/audio.h
> >>> +++ b/include/qemu/audio.h
> >>> @@ -185,6 +185,55 @@ bool audio_be_set_dbus_server(AudioBackend *be,
> >>>
> >>> const char *audio_application_name(void);
> >>>
> >>> +static inline int audio_format_bits(AudioFormat fmt)
> >>
> >> Does the inline make a difference here? My understanding is that the
> >> compiler treats this as a weak hint these days, and generally does the
> >> right thing by itself.
> >
> > If you drop "inline" you get -Werror=unused-function.
>
> Ah I guess because this is a header file? Yeah that does make it
> trickier, I guess let's keep the compiler happy then.
>
> >>
> >>> +{
> >>> + switch (fmt) {
> >>> + case AUDIO_FORMAT_S8:
> >>> + case AUDIO_FORMAT_U8:
> >>> + return 8;
> >>> +
> >>> + case AUDIO_FORMAT_S16:
> >>> + case AUDIO_FORMAT_U16:
> >>> + return 16;
> >>> +
> >>> + case AUDIO_FORMAT_F32:
> >>> + case AUDIO_FORMAT_S32:
> >>> + case AUDIO_FORMAT_U32:
> >>> + return 32;
> >>> +
> >>> + case AUDIO_FORMAT__MAX:
> >>> + break;
> >>> + }
> >>
> >> I'm not sure that AUDIO_FORMAT__MAX is a valid choice here - can you
> >> drop the explicit case for AUDIO_FORMAT__MAX and then simply have a
> >> default that calls g_assert_not_reached() instead?
> >
> > We need to handle all variants, because -Werror=switch. That
>
> Does that also include the use of default? I was thinking something
> along the lines of this to help make things clearer:
>
> static inline int audio_format_bits(AudioFormat fmt)
> {
> switch (fmt) {
> case AUDIO_FORMAT_S8:
> case AUDIO_FORMAT_U8:
> return 8;
>
> case AUDIO_FORMAT_S16:
> case AUDIO_FORMAT_U16:
> return 16;
>
> case AUDIO_FORMAT_F32:
> case AUDIO_FORMAT_S32:
> case AUDIO_FORMAT_U32:
> return 32;
>
> default:
> break;
> }
>
> g_assert_not_reached();
> }
That works, but is it any better? If we introduce a new format, it
would not warn. And if we used -Wswitch-enum, it would complain.
>
> >>> + g_assert_not_reached();
> >>> +}
> >>> +
> >>> +static inline bool audio_format_is_float(AudioFormat fmt)
> >>
> >> Same comment here re: inline.
> >>
> >>> +{
> >>> + return fmt == AUDIO_FORMAT_F32;
> >>> +}
> >>> +
> >>> +static inline bool audio_format_is_signed(AudioFormat fmt)
> >>
> >> And here.
> >>
> >>> +{
> >>> + switch (fmt) {
> >>> + case AUDIO_FORMAT_S8:
> >>> + case AUDIO_FORMAT_S16:
> >>> + case AUDIO_FORMAT_S32:
> >>> + case AUDIO_FORMAT_F32:
> >>> + return true;
> >>> +
> >>> + case AUDIO_FORMAT_U8:
> >>> + case AUDIO_FORMAT_U16:
> >>> + case AUDIO_FORMAT_U32:
> >>> + return false;
> >>> +
> >>> + case AUDIO_FORMAT__MAX:
> >>
> >> Same comment here re: AUDIO_FORMAT__MAX.
> >>
> >>> + break;
> >>> + }
> >>> +
> >>> + g_assert_not_reached();
> >>> +}
> >>> +
> >>> #define DEFINE_AUDIO_PROPERTIES(_s, _f) \
> >>> DEFINE_PROP_AUDIODEV("audiodev", _s, _f)
> >>>
> >>> diff --git a/audio/audio-mixeng-be.c b/audio/audio-mixeng-be.c
> >>> index a0e542754e5..146026d0b39 100644
> >>> --- a/audio/audio-mixeng-be.c
> >>> +++ b/audio/audio-mixeng-be.c
> >>> @@ -62,23 +62,28 @@ int audio_bug (const char *funcname, int cond)
> >>> return cond;
> >>> }
> >>>
> >>> -static inline int audio_bits_to_index (int bits)
> >>> +/*
> >>> + * Convert audio format to mixeng_clip index. Used by audio_pcm_sw_init_
> >>> and
> >>> + * audio_mixeng_backend_add_capture()
> >>> + */
> >>> +static int audio_format_to_index(AudioFormat af)
> >>> {
> >>> - switch (bits) {
> >>> - case 8:
> >>> + switch (af) {
> >>> + case AUDIO_FORMAT_U8:
> >>> + case AUDIO_FORMAT_S8:
> >>> return 0;
> >>> -
> >>> - case 16:
> >>> + case AUDIO_FORMAT_U16:
> >>> + case AUDIO_FORMAT_S16:
> >>> return 1;
> >>> -
> >>> - case 32:
> >>> + case AUDIO_FORMAT_U32:
> >>> + case AUDIO_FORMAT_S32:
> >>> return 2;
> >>> -
> >>> - default:
> >>> - audio_bug ("bits_to_index", 1);
> >>> - AUD_log (NULL, "invalid bits %d\n", bits);
> >>> - return 0;
> >>> + case AUDIO_FORMAT_F32:
> >>> + case AUDIO_FORMAT__MAX:
> >>
> >> Same comment re: AUDIO_FORMAT__MAX.
> >>
> >>> + break;
> >>> }
> >>> +
> >>> + g_assert_not_reached();
> >>> }
> >>>
> >>> void AUD_vlog (const char *cap, const char *fmt, va_list ap)
> >>> @@ -172,141 +177,68 @@ static int audio_validate_settings (const struct
> >>> audsettings *as)
> >>>
> >>> static int audio_pcm_info_eq (struct audio_pcm_info *info, const
> >>> struct audsettings *as)
> >>> {
> >>> - int bits = 8;
> >>> - bool is_signed = false, is_float = false;
> >>> -
> >>> - switch (as->fmt) {
> >>> - case AUDIO_FORMAT_S8:
> >>> - is_signed = true;
> >>> - /* fall through */
> >>> - case AUDIO_FORMAT_U8:
> >>> - break;
> >>> -
> >>> - case AUDIO_FORMAT_S16:
> >>> - is_signed = true;
> >>> - /* fall through */
> >>> - case AUDIO_FORMAT_U16:
> >>> - bits = 16;
> >>> - break;
> >>> -
> >>> - case AUDIO_FORMAT_F32:
> >>> - is_float = true;
> >>> - /* fall through */
> >>> - case AUDIO_FORMAT_S32:
> >>> - is_signed = true;
> >>> - /* fall through */
> >>> - case AUDIO_FORMAT_U32:
> >>> - bits = 32;
> >>> - break;
> >>> -
> >>> - default:
> >>> - abort();
> >>> - }
> >>> - return info->freq == as->freq
> >>> + return info->af == as->fmt
> >>> + && info->freq == as->freq
> >>> && info->nchannels == as->nchannels
> >>> - && info->is_signed == is_signed
> >>> - && info->is_float == is_float
> >>> - && info->bits == bits
> >>> && info->swap_endianness == (as->endianness !=
> >>> HOST_BIG_ENDIAN);
> >>> }
> >>>
> >>> void audio_pcm_init_info (struct audio_pcm_info *info, const struct
> >>> audsettings *as)
> >>> {
> >>> - int bits = 8, mul;
> >>> - bool is_signed = false, is_float = false;
> >>> -
> >>> - switch (as->fmt) {
> >>> - case AUDIO_FORMAT_S8:
> >>> - is_signed = true;
> >>> - /* fall through */
> >>> - case AUDIO_FORMAT_U8:
> >>> - mul = 1;
> >>> - break;
> >>> -
> >>> - case AUDIO_FORMAT_S16:
> >>> - is_signed = true;
> >>> - /* fall through */
> >>> - case AUDIO_FORMAT_U16:
> >>> - bits = 16;
> >>> - mul = 2;
> >>> - break;
> >>> -
> >>> - case AUDIO_FORMAT_F32:
> >>> - is_float = true;
> >>> - /* fall through */
> >>> - case AUDIO_FORMAT_S32:
> >>> - is_signed = true;
> >>> - /* fall through */
> >>> - case AUDIO_FORMAT_U32:
> >>> - bits = 32;
> >>> - mul = 4;
> >>> - break;
> >>> -
> >>> - default:
> >>> - abort();
> >>> - }
> >>> -
> >>> + info->af = as->fmt;
> >>> info->freq = as->freq;
> >>> - info->bits = bits;
> >>> - info->is_signed = is_signed;
> >>> - info->is_float = is_float;
> >>> info->nchannels = as->nchannels;
> >>> - info->bytes_per_frame = as->nchannels * mul;
> >>> + info->bytes_per_frame = as->nchannels * audio_format_bits(as->fmt) /
> >>> 8;
> >>> info->bytes_per_second = info->freq * info->bytes_per_frame;
> >>> info->swap_endianness = (as->endianness != HOST_BIG_ENDIAN);
> >>> }
> >>>
> >>> -void audio_pcm_info_clear_buf (struct audio_pcm_info *info, void *buf,
> >>> int len)
> >>> +void audio_pcm_info_clear_buf(struct audio_pcm_info *info, void *buf,
> >>> int len)
> >>> {
> >>> if (!len) {
> >>> return;
> >>> }
> >>>
> >>> - if (info->is_signed || info->is_float) {
> >>> - memset(buf, 0x00, len * info->bytes_per_frame);
> >>> - } else {
> >>> - switch (info->bits) {
> >>> - case 8:
> >>> - memset(buf, 0x80, len * info->bytes_per_frame);
> >>> - break;
> >>> -
> >>> - case 16:
> >>> - {
> >>> - int i;
> >>> - uint16_t *p = buf;
> >>> - short s = INT16_MAX;
> >>> -
> >>> - if (info->swap_endianness) {
> >>> - s = bswap16 (s);
> >>> - }
> >>> -
> >>> - for (i = 0; i < len * info->nchannels; i++) {
> >>> - p[i] = s;
> >>> - }
> >>> - }
> >>> - break;
> >>> + switch (info->af) {
> >>> + case AUDIO_FORMAT_U8:
> >>> + memset(buf, 0x80, len * info->bytes_per_frame);
> >>
> >> This doesn't look right - isn't this the signed version?
> >
> > No,
> >
> >>
> >>> + break;
> >>> + case AUDIO_FORMAT_U16: {
> >>> + int i;
> >>> + uint16_t *p = buf;
> >>> + short s = INT16_MAX;
> >>>
> >>> - case 32:
> >>> - {
> >>> - int i;
> >>> - uint32_t *p = buf;
> >>> - int32_t s = INT32_MAX;
> >>> + if (info->swap_endianness) {
> >>> + s = bswap16(s);
> >>> + }
> >>>
> >>> - if (info->swap_endianness) {
> >>> - s = bswap32 (s);
> >>> - }
> >>> + for (i = 0; i < len * info->nchannels; i++) {
> >>> + p[i] = s;
> >>> + }
> >>
> >> I think this is signed too?
> >>
> >>> + break;
> >>> + }
> >>> + case AUDIO_FORMAT_U32: {
> >>> + int i;
> >>> + uint32_t *p = buf;
> >>> + int32_t s = INT32_MAX;
> >>>
> >>> - for (i = 0; i < len * info->nchannels; i++) {
> >>> - p[i] = s;
> >>> - }
> >>> - }
> >>> - break;
> >>> + if (info->swap_endianness) {
> >>> + s = bswap32(s);
> >>> + }
> >>>
> >>> - default:
> >>> - AUD_log (NULL, "audio_pcm_info_clear_buf: invalid bits %d\n",
> >>> - info->bits);
> >>> - break;
> >>> + for (i = 0; i < len * info->nchannels; i++) {
> >>> + p[i] = s;
> >>> }
> >>
> >> And also here.
> >>
> >>> + break;
> >>> + }
> >>> + case AUDIO_FORMAT_S8:
> >>> + case AUDIO_FORMAT_S16:
> >>> + case AUDIO_FORMAT_S32:
> >>> + case AUDIO_FORMAT_F32:
> >>> + memset(buf, 0x00, len * info->bytes_per_frame);
> >>
> >> ... and this the unsigned version? I think they've been swapped.
> >
> > no, unsigned PCM use mid-range value for zero.
>
> Sigh. Yes of course you're right, sorry for the noise.
>
> >>> + break;
> >>> + case AUDIO_FORMAT__MAX:
> >>> + g_assert_not_reached();
> >>
> >> This looks more like I would expect, although possibly replace the case
> >> with a default? I expect that another assert() would have tripped by now
> >> with real usage if AUDIO_FORMAT__MAX were specified, but I think it
> >> would be good to be consistent.
> >
> > Unfortunately, the compiler complains for non-void functions if you
> > don't finish the function with the g_assert_not_reached().
> >
> >>
> >>> }
> >>> }
> >>>
> >>> @@ -719,8 +651,8 @@ static size_t audio_pcm_sw_write(SWVoiceOut *sw, void
> >>> *buf, size_t buf_len)
> >>> #ifdef DEBUG_AUDIO
> >>> static void audio_pcm_print_info (const char *cap, struct
> >>> audio_pcm_info *info)
> >>> {
> >>> - dolog("%s: bits %d, sign %d, float %d, freq %d, nchan %d\n",
> >>> - cap, info->bits, info->is_signed, info->is_float, info->freq,
> >>> + dolog("%s: %s, freq %d, nchan %d\n",
> >>> + cap, AudioFormat_str(info->af), info->freq,
> >>> info->nchannels);
> >>
> >> More dolog. Presumably this gets converted to error_report() later?
> >>
> >>> }
> >>> #endif
> >>> @@ -1759,15 +1691,15 @@ static CaptureVoiceOut
> >>> *audio_mixeng_backend_add_capture(
> >>>
> >>> cap->buf = g_malloc0_n(hw->mix_buf.size,
> >>> hw->info.bytes_per_frame);
> >>>
> >>> - if (hw->info.is_float) {
> >>> + if (audio_format_is_float(hw->info.af)) {
> >>> hw->clip = mixeng_clip_float[hw->info.nchannels == 2]
> >>> [hw->info.swap_endianness];
> >>> } else {
> >>> hw->clip = mixeng_clip
> >>> [hw->info.nchannels == 2]
> >>> - [hw->info.is_signed]
> >>> + [audio_format_is_signed(hw->info.af)]
> >>> [hw->info.swap_endianness]
> >>> - [audio_bits_to_index(hw->info.bits)];
> >>> + [audio_format_to_index(hw->info.af)];
> >>> }
> >>>
> >>> QLIST_INSERT_HEAD (&s->cap_head, cap, entries);
> >>> @@ -1869,29 +1801,6 @@ audsettings
> >>> audiodev_to_audsettings(AudiodevPerDirectionOptions *pdo)
> >>> };
> >>> }
> >>>
> >>> -int audioformat_bytes_per_sample(AudioFormat fmt)
> >>> -{
> >>> - switch (fmt) {
> >>> - case AUDIO_FORMAT_U8:
> >>> - case AUDIO_FORMAT_S8:
> >>> - return 1;
> >>> -
> >>> - case AUDIO_FORMAT_U16:
> >>> - case AUDIO_FORMAT_S16:
> >>> - return 2;
> >>> -
> >>> - case AUDIO_FORMAT_U32:
> >>> - case AUDIO_FORMAT_S32:
> >>> - case AUDIO_FORMAT_F32:
> >>> - return 4;
> >>> -
> >>> - case AUDIO_FORMAT__MAX:
> >>> - ;
> >>> - }
> >>> - abort();
> >>> -}
> >>> -
> >>> -
> >>> /* frames = freq * usec / 1e6 */
> >>> int audio_buffer_frames(AudiodevPerDirectionOptions *pdo,
> >>> audsettings *as, int def_usecs)
> >>> @@ -1914,8 +1823,7 @@ int
> >>> audio_buffer_samples(AudiodevPerDirectionOptions *pdo,
> >>> int audio_buffer_bytes(AudiodevPerDirectionOptions *pdo,
> >>> audsettings *as, int def_usecs)
> >>> {
> >>> - return audio_buffer_samples(pdo, as, def_usecs) *
> >>> - audioformat_bytes_per_sample(as->fmt);
> >>> + return audio_buffer_samples(pdo, as, def_usecs) *
> >>> audio_format_bits(as->fmt) / 8;
> >>> }
> >>>
> >>> void audio_rate_start(RateCtl *rate)
> >>> diff --git a/audio/dbusaudio.c b/audio/dbusaudio.c
> >>> index e284542b2dd..72d6194033b 100644
> >>> --- a/audio/dbusaudio.c
> >>> +++ b/audio/dbusaudio.c
> >>> @@ -147,9 +147,9 @@
> >>> dbus_init_out_listener(QemuDBusDisplay1AudioOutListener *listener,
> >>> qemu_dbus_display1_audio_out_listener_call_init(
> >>> listener,
> >>> (uintptr_t)hw,
> >>> - hw->info.bits,
> >>> - hw->info.is_signed,
> >>> - hw->info.is_float,
> >>> + audio_format_bits(hw->info.af),
> >>> + audio_format_is_signed(hw->info.af),
> >>> + audio_format_is_float(hw->info.af),
> >>> hw->info.freq,
> >>> hw->info.nchannels,
> >>> hw->info.bytes_per_frame,
> >>> @@ -273,9 +273,9 @@ dbus_init_in_listener(QemuDBusDisplay1AudioInListener
> >>> *listener, HWVoiceIn *hw)
> >>> qemu_dbus_display1_audio_in_listener_call_init(
> >>> listener,
> >>> (uintptr_t)hw,
> >>> - hw->info.bits,
> >>> - hw->info.is_signed,
> >>> - hw->info.is_float,
> >>> + audio_format_bits(hw->info.af),
> >>> + audio_format_is_signed(hw->info.af),
> >>> + audio_format_is_float(hw->info.af),
> >>> hw->info.freq,
> >>> hw->info.nchannels,
> >>> hw->info.bytes_per_frame,
> >>> diff --git a/audio/coreaudio.m b/audio/coreaudio.m
> >>> index 40d7986b1d7..08bab353831 100644
> >>> --- a/audio/coreaudio.m
> >>> +++ b/audio/coreaudio.m
> >>> @@ -359,7 +359,7 @@ static OSStatus init_out_device(coreaudioVoiceOut
> >>> *core)
> >>> AudioValueRange frameRange;
> >>>
> >>> AudioStreamBasicDescription streamBasicDescription = {
> >>> - .mBitsPerChannel = core->hw.info.bits,
> >>> + .mBitsPerChannel = audio_format_bits(core->hw.info.af),
> >>> .mBytesPerFrame = core->hw.info.bytes_per_frame,
> >>> .mBytesPerPacket = core->hw.info.bytes_per_frame,
> >>> .mChannelsPerFrame = core->hw.info.nchannels,
>
> Having said that, given that the unsigned/signed bit is actually
> correct, then regardless of the final form of the switch() statements
> then this seems okay:
>
> Reviewed-by: Mark Cave-Ayland <[email protected]>
>
>
thnks
--
Marc-André Lureau
