Fwd: tried to make giflist, got segfault
Attached is the Freelancer patch so if Andrew and Andrea ould test (and Andrew look at the code) to see if it is usable, I would appreciate it. (Andrea, cd to ...cinelerra-5.1/cinelerra and do "patch < location of downloaded patch" should work. On Mon, Dec 13, 2021 at 12:07 AM Andrew Randrianasulu < [email protected]> wrote:
On Sunday, December 12, 2021, Phyllis Smith <[email protected]> wrote:
this patch prevent gif segfault problem (but all images become the same for given list!) , but not sure about other, similar crashes you observed?
Andrew, patch works just like you stated and most importantly prevents the crash. Similar mod would have to be put into fileexr.C, filejpeg.C, etc. but let me see if I can get the Freelancer to come up with a different solution -- he keeps nagging me for something to do.
well, I think real solution should augment those filereaders with something like report_w, report_h and then caller must check if previous (cached) info remain the same for current frame.. so at first call we check those params and load them into var1_w, and var1_h, then set someflag we_read_first_info, then just copy var1_w/h to var2_w/h, repeat, now when we have someflag set (thus not first read) compare those two, abort reading of mismatched frame.
after trying to write down it like this I even think we can name vars first_w, first_h and then check current fetched values against them, not re-loading pair for next/priv frame at each step. this is for avoiding quite expensive decoding, if I understand our code correctly!
good luck to freelancer, good health to both of you...
I agree that it needs a safeguard. Let me know if you do not find a
solution because a freelancer wants to work on something and I thought he might be able to actually do this. Most of the problems are way too difficult for someone not familiar with Cinelerra to even attempt to work on.
Sorry for the delay. I compiled with the patch with no problems. I tried using GIF Squence (on 2 edits in a track) which produced a series of images with a toc. Importing the toc into CinGG produces a working asset. Playing the render on the timeline shows that the first edit rendered correctly; the second has the colors wrong (posterization). Then I tried rendering with ffmpeg --> git. I got a single animated gif image (and not many images with a toc). In this case both rendered edits have false colors in timeline. Reading the image sequences is usually smooth on the timeline, and in fact in the first case it's smooth; instead in the second case it's problematic with temporary freezes. No message on the terminal.
Andrea, the second has the colors wrong (posterization).
Then I tried rendering with ffmpeg --> git. I got a single animated gif image (and not many images with a toc). In this case both rendered edits have false colors in timeline. Reading the image sequences is usually smooth on the timeline, and in fact in the first case it's smooth; instead in the second case it's problematic with temporary freezes. No message on the terminal.
I think you might have found an additional problem. Without the patch in, do you still get the false colors using the same scenario?
Without the patch in, do you still get the false colors using the same scenario? Yes, same problems even without patch. It's hard for me to understand if they are problems with my system or with some patch introduced lately. There are 3 problems that I currently have: 1- Mask not working in OpenGL 2- Wrong colors in rendering image sequences 3- Transparency (Alpha) issues with the Title plugin seen in BT#555 and #600.
participants (2)
-
Andrea paz -
Phyllis Smith