tinyalsa: fix: format conversion

[gbarkadiusz]

Expand ALSA text to bit depth conversion function with support for FLOAT_LE, U8, MU_LAW, and A_LAW.

Justification.
The check-capture test uses format that doesn't contain the digit. " e.g., FLOAT_LE".
As a result, the converting function returns an empty string, which the shell interprets as 0. This causes the tinycap command to fail with the following error:
tinycap /dev/null -D 0 -d 0 -c 2 -t 3 -r 48000 -b 0 bits is not supported.

[golowanow]

there are other format strings now also possible, e.g.

Device Information
        Name: hw:sofnocodec,0
        Card: sofnocodec[sof-nocodec]
        Device: Port0 (*)[]
        Stream: CAPTURE
        Format: ['U8', 'S16_LE', 'S24_LE', 'S32_LE', 'FLOAT_LE', 'MU_LAW', 'A_LAW']

[golowanow]

the above is just an example from one configuration, but there could be other
https://www.alsa-project.org/alsa-doc/alsa-lib/group___p_c_m.html#gaa14b7f26877a812acbb39811364177f8

ie. currently this test can handle '[0-9]+_BE`, but with this change it will not.

[golowanow]

I think _[SU]([0-9])+_ and _FLOAT([0-9]+)_ should cover many formats and the rest (like 'FLOAT', '*_LAW') can be hardcoded.

[gbarkadiusz]

So how about this one:

  case "$format_string" in
        # 8-bit formats (Signed/Unsigned)
        S8 | U8)
            printf '8'
            ;;
        # 16-bit formats (Signed/Unsigned, various endianness)
        S16_LE | S16_BE | U16_LE | U16_BE | S16 | U16)
            printf '16'
            ;;
        # 18-bit formats (3-byte container, various endianness)
        S18_3LE | S18_3BE | U18_3LE | U18_3BE)
            printf '18'
            ;;
        # 20-bit formats (in 4-byte or 3-byte containers, various endianness)
        S20_LE | S20_BE | U20_LE | U20_BE | S20 | U20 | \
        S20_3LE | S20_3BE | U20_3LE | U20_3BE)
            printf '20'
            ;;
        # 24-bit formats (in 4-byte or 3-byte containers, various endianness)
        S24_LE | S24_BE | U24_LE | U24_BE | S24 | U24 | \
        S24_3LE | S24_3BE | U24_3LE | U24_3BE)
            printf '24'
            ;;
        # 32-bit integer formats (Signed/Unsigned, various endianness)
        S32_LE | S32_BE | U32_LE | U32_BE | S32 | U32)
            printf '32'
            ;;
        # 32-bit Floating-point formats
        FLOAT_LE | FLOAT_BE | FLOAT)
            printf '32' # Standard single-precision float
            ;;
        # 64-bit Floating-point formats
        FLOAT64_LE | FLOAT64_BE | FLOAT64)
            printf '64' # Double-precision float
            ;;
        # Compressed / Encoded formats (typically handled as 8-bit for raw stream)
        MU_LAW | A_LAW | IMA_ADPCM | MPEG | GSM)
            printf '8' 
            ;;
        # IEC958 (S/PDIF) subframe formats
        IEC958_SUBFRAME_LE | IEC958_SUBFRAME_BE | IEC958_SUBFRAME)
            printf '24' # IEC958 commonly carries up to 24-bit audio data
            ;;
        # Formats without a direct, standard bit depth mapping or needing special handling
        UNKNOWN | SPECIAL) # Assuming 'UNKNOWN' corresponds to SND_PCM_FORMAT_UNKNOWN
            printf '' # No direct bit depth, or requires specific codec/driver handling
            ;;
        # Default case for any unlisted or unrecognized formats
        *)
            # echo "ERROR: Unrecognized format encountered: $format_string" >&2
            printf '' 
            ;;
    esac

[golowanow]

please check against https://github.com/alsa-project/alsa-lib/blob/master/src/pcm/pcm_misc.c#L210
and squeeze it with patterns following the naming convention, whether possible.

[ujfalusi]

Just a note: we are never going to encounter most of these formats (GSM, U/S18, *_BE)

I'm uncertain what is the meaning of bits in tiny-world...
S24_LE is 24bit audio in 32bit container, LSB aligned
S24_3LE s 24bit audio in 24bit container (sort of packed, no padding)
S32_LE + MSBITS_24 is 24bit audio in 32bit container, MSB aligned (we don't yet have support for this in SOF)

which of these -b24 will result? My guess is S24_LE, but I'm not sure.

[gbarkadiusz]

Assuming that, I've confirmed that expanding the supported format list to allow FLOAT makes the test pass.

[golowanow]

Just a note: we are never going to encounter most of these formats (GSM, U/S18, *_BE)

Let's implement it as explicit case selection for only these formats we use for validation.

[golowanow]

please extend your commit's summary with more details, e.g.
Expand ALSA text to bit size format conversion function with FLOAT_LE and ....

[golowanow]

SOFCI TEST

[golowanow]

Would be nice to update the commit message to what is actually done with refactoring other format's handling in addition to FLOAT.