The bug could be due to ffmpeg 5.1 as opposed to ffmpeg 4.x applied patches or some other change. Valgrind (heaptrack is really great but I am too lazy to git it) on Oct 31, 2020 does not exhibit this astronomical loss of available memory. October 31, 2020 GIT ==257884== definitely lost: 256 bytes in 8 blocks ==257884== *indirectly lost: 64 bytes* in 8 blocks ==257884== possibly lost: 0 bytes in 0 blocks Current GIT: ==4985== definitely lost: 1,296 bytes in 9 blocks ==4985== * indirectly lost: 1,626,411 bytes* in 272 blocks ==4985== possibly lost: 528,288 bytes in 14 blocks On Mon, Jan 23, 2023 at 3:01 PM Stefan de Konink <[email protected]> wrote:
On Monday, January 23, 2023 10:56:00 PM CET, Andrew Randrianasulu wrote:
sorry, I can't think about anything BUT disabling it by early return in functions allocating cache items ...
I added a -1, as effective.
Lets say the solution is the file should be closed, but I am not yet at the point what either "does not do that" (mwindow/vwindow) or "is logically wrong" (garbage collector).
-- Stefan