On Monday, May 24, 2021, Phyllis Smith via Cin <cin@lists.cinelerra-gg.org> wrote:

Andrew,
try this patch?

Thanks for the patch - I should have known the fix went into ffmpeg.C instead of fileffmpeg.C. Sorry for getting it wrong.

BUT, the bad news is that the patch + uncommenting the ff->video_probe(1); line in fileffmpeg.C does not solve the
use of extra memory. It seemed like it might have though. I finally found a definitive test case and this time saved it
so I could find it again.

cool!

Bummer. Since I often make mistakes, I ran it several times and verified that I correctly had the lines right. The
results were consistent -- as soon as I uncommented "ff->video_probe(1); in fileffmpeg.C" whether or not the patch
to ffmpeg.C was in, the memory usage went up about 18%.

but does this memory usage increase with each loaded clip/media, or stay the same? does memory actually leak (usage grow, and some memory still not freed after program quit)?

18% sounds a lot (too much for single-frame decoding i was hoping for)

On Mon, May 3, 2021 at 10:09 PM Andrew Randrianasulu via Cin <cin@lists.cinelerra-gg.org> wrote:
I think I spotted my error, resulting in memory leak..

Try to add

av_frame_free(&frame);

Just before return ret; line.. { line 1246 currently?}

And after this re-enable line disabled in
commit eabda9cf447580a974fd3d67a3d0a26f09aeb523

ff->video_probe(1); in fileffmpeg.C

And of course re-test ( i probably loaded only very small project or number of files for noticing this.. :/)