While working on kew I ran into a few problems that I fixed in miniaudio.h. I'm making this PR to provide you with these fixes. You don't need to include them of course, but it would make life easier for us, because our package managers would be able to use miniaudio as an external dependency instead of vendoring without issue.

The fixes included are:

1. call to MultiByteToWideChar:
   sizeof(filenameW) returns the total size of the buffer in bytes (4096 bytes)
   but MultiByteToWideChar is expecting wide characters (wchar_t), not bytes.
   Solution: divide by sizeof(WCHAR)

2. Added read error check in static ma_result ma_device_stop__alsa(ma_device* pDevice)
   The original code checks if the number of bytes read (resultRead) is not equal to the expected number of bytes (sizeof(t))
   However, it doesn't handle the possibility of an actual read() error. read() can return -1. This check has been added with a call to perror. This might not be exactly how you'd like to handle it, but I have included it in any case.

3. Added default value
   ma_uint64 framesProcessed;
   changed to:
   ma_uint64 framesProcessed = 0;
   in MA_API ma_result ma_data_source_read_pcm_frames(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead)

4. Fix potential division by 0 error
   int frame_bytes = ma_dr_mp3_hdr_frame_samples(h)*ma_dr_mp3_hdr_bitrate_kbps(h)*125/ma_dr_mp3_hdr_sample_rate_hz(h);
   changed to:
   unsigned int sampleRate = ma_dr_mp3_hdr_sample_rate_hz(h);
   if (sampleRate == 0) return 0;
   int frame_bytes = ma_dr_mp3_hdr_frame_samples(h)*ma_dr_mp3_hdr_bitrate_kbps(h)*125/sampleRate;

5. Return if userdata is null
   if (pMP3->pUserData == NULL)
   return 0;
   added to static size_t ma_dr_mp3__on_read(ma_dr_mp3* pMP3, void* pBufferOut, size_t bytesToRead)

That's it!