SIGFPE on certain audio files

[tazz4843]

Hey there! I'm testing out whisper.cpp to see if it would be suitable for production use. However I'm running into a SIGFPE on certain audio files: namely those that do not produce any output from the model. Because of the way my system is set up, I'm unable to provide any test files that can reproduce this bug.

However, I was able to build the library with debug symbols and trigger the exception. It seems to be a divide-by-zero error on line 2349 of whisper.cpp:

whisper.cpp/whisper.cpp

Line 2349 in 8d94358

int progress_cur = (100*seek)/whisper_n_len(ctx);

The GDB output is as follows:

Thread 21 "scripty_stt_ser" received signal SIGFPE, Arithmetic exception.
[Switching to Thread 0x7ffff7085700 (LWP 3869)]
0x0000555555599123 in whisper_full (ctx=0x5555556f6a80, params=..., samples=<optimized out>, n_samples=<optimized out>) at whisper.cpp:2349
2349            int progress_cur = (100*seek)/whisper_n_len(ctx);

Unfortunately, despite compiling with debug symbols (-g flag), bt gave no extra info beyond that:

(gdb) bt
#0  0x0000555555599123 in whisper_full (ctx=0x5555556f6a80, params=..., samples=<optimized out>, n_samples=<optimized out>) at whisper.cpp:2349
#1  0x0000555555593cf6 in whisper_rs::whisper_ctx::WhisperContext::full (self=<optimized out>, params=..., data=...) at src/whisper_ctx.rs:390

Let me know if there's anything else I can do to help!

[ggerganov]

Very likely that the length of the audio is 0.
The function whisper_n_len returns the length of the spectrogram and it will be 0 if the audio is empty.

There should be a check for division by zero.

When using -g you also want to avoid using -O3 in order to not get <optimized out>

[tazz4843]

I'm willing to try out implementing this, how would you want the division by zero check to be handled? ie. should it throw an error or should it just silently skip?

[ggerganov]

Right after computing the mel spectrogram, check if the length is less than 100 (i.e. 1 second) and if yes - return 0:

whisper.cpp/whisper.cpp

Lines 2315 to 2320 in 8d94358

	// compute log mel spectrogram
	if (whisper_pcm_to_mel(ctx, samples, n_samples, params.n_threads) != 0) {
	fprintf(stderr, "%s: failed to compute log mel spectrogram\n", __func__);
	return -1;
	}

Add short comment with explanation that we do not process audio less than 1 second