Martin Guy <martinw...@gmail.com> writes:
> On 5/20/24 15:27, Måns Rullgård wrote:
>> Martin Guy <martinw...@gmail.com> writes:
>>> I would post an issue and patch on sourceforge, but that seems useless as
>>> the existing patches haven't been applied since 2006.
>> You can send patches to this mailing list.
>
> Here.
>
> In general, though, sox's SF issue tracker seems to be ineffective, which
> both impacts the quality and functionality of sox, but also discourages
> well-intentioned people from contributing to it, as they start out
> enthusiastic, follow what seem like the right steps, and then are ignored
> for decades. I am one example of that. I've added log frequency axis to
> sndfile-tools' sndfile-spectrogram and would have done the same for sox but
> it seemed pointless after the FFTW patches being ignored for so long, making
> me have to update them to stride code changes, until I eventually gave the
> idea up as wasted work.
>
> There are contributions with patches from 2006, which may or may not be
> useful these days. There are patch suggestions that give links to github
> repos that have long since been deleted, but the pointless issue persists on
> SF.
>
> This gives the public impression to developers is that it is abandonware or
> only-us-not-you-ware.
>
> Could the project use some help, as I've been programming since I was 13 and
> am now 60 and on a pension so I can choose what I work on, and resolving the
> sox SF issues seems worth doing as it is an excellent tool that seems
> neglected in that area.
>
> Cheers
>
> M
>
> Fix segmentation fault when -x argument is > number of samples in the file
>
> $ sox -c 1 -b 16 -r 1000 -n file.wav synth 1 sin 440 gain -3
> $ sox file.wav -n spectrogram -x 1001
> Segmentation fault (core dumped)
>
> #0 0xb7e77c29 in print_at_ (pixels=0xa5e630 "", cols=144, x=-2147483614,
> y=24, c=1, text=0xbfe4b824 "Time (s)", orientation=0) at spectrogram.c:728
> 728 case 0: pixel(x + j, y - i) = c; break;
> (gdb) frame 1
> #1 0xb7e78749 in stop (effp=0xa46410) at spectrogram.c:870
> 870 print_at(left + (p->cols - font_X * strlen(text)) / 2, 24, Text,
> text);
>
> This happens because strlen(txt) is unsigned, making the whole bracket
> expression unsigned and calculating 4294967248/2 instead of -48/2
>
> diff --git a/src/spectrogram.c b/src/spectrogram.c
> index 3dcda69c..c13eb6c9 100644
> --- a/src/spectrogram.c
> +++ b/src/spectrogram.c
> @@ -867,7 +867,7 @@ static int stop(sox_effect_t *effp) /* only called, by
> end(), on flow 0 */
> /* X-axis */
> step = axis(secs(p->cols), p->cols / (font_X * 9 / 2), &limit, &prefix);
> sprintf(text, "Time (%.1ss)", prefix); /* Axis label */
> - print_at(left + (p->cols - font_X * strlen(text)) / 2, 24, Text, text);
> + print_at(left + (p->cols - font_X * (int)strlen(text)) / 2, 24, Text,
> text);
>
> for (i = 0; i <= limit; i += step) {
> int x = limit ? (double)i / limit * p->cols + .5 : 0;
There must be something else/additional going on, because on my system
it segfaults with or without this patch.
--
Måns Rullgård
_______________________________________________
SoX-devel mailing list
SoX-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sox-devel
