error Assertion `nout <= freedv_get_n_max_speech_samples(f)' failed.

[netmate1234]

freedv: /home/martin/temp/freedv-gui/codec2/src/freedv_api.c:823: freedv_bits_to_speech: Assertion `nout <= freedv_get_n_max_speech_samples(f)' failed. Aborted (core dumped)

(Running 2020. It comes often, Ubuntu 20.10. FreeDV 1.5.2 dev.)

[drowe67]

Pls see https://github.com/drowe67/freedv-gui/issues/107 suggestions (1) and (2)

[drowe67]

Closed as no response from OP

[netmate1234]

I removed 3 assserts in codec2. now its stable.

[drowe67]

Removing asserts is a bad idea - they are there to trap unexpected and illegal behavior.

[drowe67]

@tmiw :point_up: this might be related to what Kanda is reporting :thinking:

[drowe67]

@timw I popped this into the top of freedv_rx and the assert fired:

    fprintf(stderr, "max_speech_samples: %d max_modem_samples: %d\n", 
        freedv_get_n_max_speech_samples(freedv), freedv_get_n_max_modem_samples(freedv));
    {
        int nout = 2*freedv_get_n_max_modem_samples(freedv);
        assert(nout <= freedv_get_n_max_speech_samples(freedv));
    }

I think I know what's going on here:

1. When we are out out sync (and squelch is off), we run through that pass through code, the input modem samples are oversampled up to the 16 kHz speech sample rate so we can hear the off air rx audio.
2. When the demod is out of sync it will just ask for the nominal number of input samples, which satisfies the assert.
3. However if we fall out of sync, it's possible nin might be set to max_modem_samples on the last modem frame processed before we fall out of sync. Doesn't look like there is enough storage for that case, so the assert fires.
4. I haven't been able to reproduce the actual issue, as it requires a few conditions to line up, but can see how it might happen from analysis. Just in 2020 I suspect.

[drowe67]

OK by forcing nin to be the max in freedv_rx.c:

        nin = freedv_nin(freedv);
        nin = freedv_get_n_max_modem_samples(freedv);
        /* optionally read some stats */

I can also trigger another assert (I have seen this once before):

./src/freedv_tx 2020 ~/LPCNet/wav/all.wav - | ./src/cohpsk_ch - - -20 | ./src/freedv_rx 2020 - /dev/null -v
<snip>
freedv_rx: /home/david/codec2/src/fdmdv.c:1770: fdmdv_8_to_16: Assertion `(n % FDMDV_OS) == 0' failed.

I hit that once before.

[tmiw]

@drowe67, I assume the PR fixes that additional assert as well?

[drowe67]

I think this one was fixed with https://github.com/drowe67/codec2/pull/176