fix(mod_shout): update handle->pos in shout_file_read() to fix backward seek for MP3 files

[pinc444]

mod_shout's shout_file_read() never updated handle->pos, causing switch_core_file_seek() to always compute 0 - samps for backward seeks — clamping to position 0 instead of the correct relative position.

Change

Added handle->pos += *len in shout_file_read(), matching the pattern already used by mod_sndfile:

// src/mod/formats/mod_shout/mod_shout.c — shout_file_read()
handle->pos += *len;          // was missing; mod_sndfile has this
handle->sample_count += *len;

This ensures fh->pos accurately reflects playback position so that callers doing fh->pos - samps (backward seek via SEEK_SET) compute the correct target sample.

Original prompt

Bug

When using session.streamFile() with an MP3 file (handled by mod_shout), backward seek does not work correctly. The file always seeks to position 0 instead of the expected relative backward position.

WAV files work correctly because mod_sndfile properly tracks handle->pos in its file_read() function. mod_shout does not update handle->pos during reads, which breaks the seek logic in switch_core_file_seek().

Root Cause

In src/mod/formats/mod_shout/mod_shout.c, the function shout_file_read() increments handle->sample_count but never updates handle->pos.

Meanwhile, in src/switch_core_file.c, the switch_core_file_seek() function uses fh->offset_pos for SEEK_CUR operations:

if (whence == SWITCH_SEEK_CUR) {
    unsigned int cur = 0;
    if (switch_test_flag(fh, SWITCH_FILE_FLAG_WRITE)) {
        fh->file_interface->file_seek(fh, &cur, fh->samples_out, SEEK_SET);
    } else {
        fh->file_interface->file_seek(fh, &cur, fh->offset_pos, SEEK_SET);
    }
}

And fh->offset_pos is set from *cur_pos which comes from the file_seek return. But the first seek (to establish current position) relies on fh->offset_pos being correct. Since mod_shout never updates handle->pos (and offset_pos is derived from it after seeks), the position is always 0, making backward seeks always go to position 0.

For comparison, mod_sndfile in src/mod/formats/mod_sndfile/mod_sndfile.c correctly updates handle->pos in its read function:

static switch_status_t sndfile_file_read(switch_file_handle_t *handle, void *data, size_t *len)
{
    // ... read logic ...
    handle->pos += *len;        // <-- THIS IS WHAT mod_shout IS MISSING
    handle->sample_count += *len;
    return *len ? SWITCH_STATUS_SUCCESS : SWITCH_STATUS_FALSE;
}

Fix Required

In src/mod/formats/mod_shout/mod_shout.c, find the shout_file_read() function. It currently has a line:

handle->sample_count += *len;

Add the following line immediately before that existing line:

handle->pos += *len;

So the result looks like:

handle->pos += *len;
handle->sample_count += *len;

This is the only change needed. It matches the pattern used by mod_sndfile and ensures that switch_core_file_seek() has the correct current position when computing relative seeks (both forward and backward via SEEK_CUR).

The following is the prior conversation context from the user's chat exploration (may be truncated):

Assistant:

# Generates a 2-minute WAV file with a voice announcing each second
# Uses .NET System.Speech (built into Windows)

Add-Type -AssemblyName System.Speech

$SAMPLE_RATE = 8000
$OUTPUT = Join-Path (Get-Location) "counting_seconds.wav"
$TEMP_DIR = Join-Path $env:TEMP "counting_$(Get-Random)"
New-Item -ItemType Directory -Path $TEMP_DIR -Force | Out-Null

$TOTAL_SECONDS = 120

Write-Host "Generating $TOTAL_SECONDS seconds of spoken counting audio..."
Write-Host ""

$synth = New-Object System.Speech.Synthesis.SpeechSynthesizer

for ($i = 0; $i -le $TOTAL_SECONDS; $i++) {
    $voiceFile = Join-Path $TEMP_DIR "voice_$($i.ToString('D4')).wav"
    $finalFile = Join-Path $TEMP_DIR "final_$($i.ToString('D4')).wav"

    # Generate spoken number to WAV
    $synth.SetOutputToWaveFile($voiceFile)
    $synth.Rate = 3  # faster speech so it fits in 1 second
    $synth.Speak("$i")
    $synth.SetOutputToNull()

    # Convert to 8kHz mono and pad/trim to exactly 1 second
    $soxArgs = @(
        "`"$voiceFile`""
        "-r", "$SAMPLE_RATE"
        "-c", "1"
        "`"$finalFile`""
        "pad", "0", "1.0"
        "trim", "0", "1.0"
    )
    Start-Process -FilePath "sox" -ArgumentList ($soxArgs -join " ") -NoNewWindow -Wait

    Write-Host "`r  Generated second $i / $TOTAL_SECONDS" -NoNewline
}

$synth.Dispose()

Write-Host ""
Write-Host "Combining all seconds into $OUTPUT..."

# Get all final files in order
$allFiles = (0..$TOTAL_SECONDS | ForEach-Object {
    "`"$(Join-Path $TEMP_DIR "final_$($_.ToString('D4')).wav")`""
}) -join " "

Start-Process -FilePath "sox" -ArgumentList "$allFiles `"$OUTPUT`"" -NoNewWindow -Wait

Write-Host ""
Write-Host "=========================================="
Write-Host "  FILE GENERATED: $OUTPUT"
Write-Host "=========================================="
Write-Host ""
Write-Host "  You will hear: 'zero' 'one' 'two' ... 'one hundred twenty'"
Write-Host "  Each number takes exactly 1 second."
Write-Host "  The number you hear = the second position in the file."
Write-Host ""
Write-Host "HOW TO TEST:"
Write-Host ""
Write-Host "  1. Play this file with your streamFile + onPlayFile callback"
Write-Host "  2. Wait until you hear 'sixty'"
Write-Host "  3. Press BACKWARD key (1) - seek:-15000"
Write-Host "     Should hear: 'forty-five' (60 - 15 = 45)"
Write-Host "     Bug: you hear 'zero' or 'one'"
Write-Host ""
Write-Host "  4. Reset. Wait until you hear 'sixty' again"
Write-Host ...

</details>