November 2022 - Cin - Cinelerra-GG Mailing Lists

Fwd: more dynamic libs for cinelerra/Makefile
by Andrew Randrianasulu 19 Nov '22

19 Nov '22

I strongly apologize, but apparently my last shared build missed those libdv-related libs. Now shared build running again (with LIBGL_DRI3_DISABLE=1 even needed on r600 driver) but I can't see crash with changing project's format between rgb-8 and rgba-8 ? Can you point at specific plugins you loaded? ---------- Forwarded message --------- От: Андрей Рандрианасулу <randrik(a)mail.ru> Date: пт, 18 нояб. 2022 г., 07:31 Subject: more dynamic libs for cinelerra/Makefile To: randrianasulu <randrianasulu(a)gmail.com> -- Андрей Рандрианасулу

2 2

Reproducible crash with specific files and steps
by Phyllis Smith 19 Nov '22

19 Nov '22

Here are the location of the 2 files and the exact steps to create a reproducible crash ON MY LAPTOP (so maybe it is me?). Wondering if Andrea can reproduce on Arch since he always keeps his system up to date? Input files at: https://streamable.com/ppm3hx https://streamable.com/ft4qhb Exact steps: 1) Startup CinGG from a window using new bcast files like: CIN_CONFIG=/tmp/bcast1 bin/cin 2) Load file of rainbow colors (from ppm3hx streamable) using strategy "replace current project". Click off error message. 3) Load file of flowers (from ft4qhb streamable) using strategy "append in new tracks". 4) Using the Video pulldown, choose "attach effect" and pick the Overlay effect. 5) Check the Settings pulldown and see that Format is currently at RGBA-8; switch to RGB-8. 6) Type "q" for quit in the main program window and choose "No". 7) Now run again using the same startup type of: CIN_CONFIG=/tmp/bcast1 bin/cin 8) Do not do anything else, just use the File pulldown and choose "Load backup". You can X the Tip window. 9) Using the Settings pulldown, choose Format and switch to RGBA-8. 10) Wait for it -- Boom, it crashes and any time you try to Load backup, it crashes.

3 8

OT: libquicktime in AUR
by Andrew Randrianasulu 18 Nov '22

18 Nov '22

Andrea, you are Arch user, if you registered at their bug tracing system related to AUR can you look into this entry? https://aur.archlinux.org/packages/libquicktime see discussion, it also must be noted this July Debian maintainer posted untested ffmpeg5 patch to Debian BTS but it was decided to disable ffmpeg support altogether ... https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1004637 because I used lqtplay literally today while working on some cin-cv code ... I do not want this codebase die ..... can you post link to debian patch there? https://bugs.debian.org/cgi-bin/bugreport.cgi?att=1;bug=1004637;filename=24… and if possible test it? (my main system have ffmpeg 4.4.3 and termux might have some additional weirdness not accounted for)

2 1

Fwd: cingg: possible crash with undo on nested session
by Andrew Randrianasulu 17 Nov '22

17 Nov '22

---------- Forwarded message --------- От: Андрей Рандрианасулу <randrik(a)mail.ru> Date: вс, 13 нояб. 2022 г., 01:35 Subject: cingg: possible crash with undo on nested session To: randrianasulu <randrianasulu(a)gmail.com> Steps: Load 1 video, 2 audio tracks video into Cingg with Overwrite current project mode Modify video fade slider down Save xml for future nested use Open another image, so it will overwrite proj to single video track Add video track Open previously saved xml as nested , in resources only. Drag nested edl to top empty video track (where is alpha?) click once on ‘output this track’ toggle on bottom video track's pathbay click on same toggle in pathbay second time. Undo (‘z’) at this point it crashed for me I’ll attach three patches I made trying to debug this — they do not fix THE crash but make backtrace less scary gdb.txt also attached -- Андрей Рандрианасулу

2 1

some cleanups, adding ffmpeg script from mail archive
by Andrew Randrianasulu 14 Nov '22

14 Nov '22

Using old thread https://lists.cinelerra-gg.org/pipermail/cin/2019-May/000671.html I added vaapi proxying script to our doc/ directory (really not fan of retyping scripts from manual ...) and removed near-empty "whats new" from 2.1 era and removed few 'sav' backup copies from tree and moved old 1.1.0 release notes and TODOs from svn era into new old/ directory. see if you like those changes, apply accordingly!

3 5

Fwd: beter fix for nested edl load/undo
by Andrew Randrianasulu 13 Nov '22

13 Nov '22

oh, hopefully correct fix this time I was hitting strange bug when undoing just loaded nested edl was causing segfault. After MUCH mucking around I think I discovered we were calling wrong function via pointer, I tested loading/undoing nested edl, normal edl, reference edl and group-loading 25+ pngs on separate tracks or as 25 pieces on same track. And undoing a lot! But please test as hard as possible - threading is not my strong suit, actually it is not my suit at all! ---------- Forwarded message --------- От: Андрей Рандрианасулу <randrik(a)mail.ru> Date: вс, 13 нояб. 2022 г., 15:20 Subject: beter fix for nested edl load/uno To: randrianasulu <randrianasulu(a)gmail.com> -- Андрей Рандрианасулу

1 0

Cingg'divide mode, very divisive!
by Andrew Randrianasulu 10 Nov '22

10 Nov '22

So, as Igor noticed CinGG's divide and subtract modes do not behave like they behaved in Cin HV/CV/CVE I digged into matter a bit and found few confusing (for me, amateur) things. First, it seems CinGG as of now implement divide mode just like Natron (2.5 from git). https://paste.pics/JJC90 But GIMP/Krita/Photoshop implement different algo, probably described here: https://github.com/psd-tools/image-blender/blob/master/wiki/Blend-modes-and… basically it makes color white if one input (not one you divide to) was black. I think this is consistent with screenshot: https://paste.pics/JJC76 But moreover, according to documentation CinGG composites tracks a bit backward comparing to photo editors: Description The previous math forms are the only truly accurate description of each blending operation, but short descriptions are below where Source is the output from the next track and Destination is the output from the lower track stacking. Blending starts with new Source and combines it with the previous render stack output, which is referred to as Destination. The new output becomes the next Destination and the next up stack level becomes the new Source. Source is above; Destination is below. --- https://cinelerra-gg.org/download/CinelerraGG_Manual/Description.html So, in subtract and divide modes this *really* makes a lot of difference, like inverting (wildly altering) colors. By playing with automatically set keyframes I 'discovered' you can change compositing mode along the track - so this makes interesting dynamic effects possible. But I do not think reverting subtract/divide to their original form will be best solution. I hope another set of modes can be added, implementing at least previous algo, even if not exact ordering of compositing. (you always can swap tracks or use overlay plugin for different order?) I did similar (software-only ) patch ~year back, attached again (might need some tweaking for new sources). Problem is, I am a bit lost in this MACRO thing, so showing everyone how to put complex equation into our current overlayframe.h very much hoped for. Of course more ultimate idea is to put user-editable inputbox somewhere there, yet I not dare to think what will happen if user just added some A/B and B happened to be zero ..... at runtime, due to picture (so, macro somewhat must wrap around existing checking macro, and THIS definitely above my level ) https://git.cinelerra-gg.org/git/?p=goodguy/cinelerra.git;a=blob;f=cinelerr… One the bonus side, I made simple PSD file in GIMP (2.10.32) and it was openable in cingg via ffmpeg's decoder! of course I am not sure if ffmpeg ever will expose raw layers and metadata for making more complex composition work in applications using it possible, but hey, who knows the future! attached even screenshot from psd-tools repo because I am a bit paranoid about information disappearance ....

2 4

Mailing list - file upload
by Sam 10 Nov '22

10 Nov '22

Hi guys, I am very pleased that the mailing list is used by our visitors and users to exchange information, that should remain so, but please do not send too large files over the mailing list, it is not intended for that. This causes us more administrative work and higher costs for storage space on the server. Larger files should be shared via file sharing services and included as a link in the email. For larger files, swisstransfer [dot] com (high Swiss data protection law) is a good choice, or streamable [dot] com, or for images imgbb [dot] com, etc. For this reason, e-mails that are too large will be automatically rejected. Thank you for your understanding. Sam

1 0

opencl on Radeons
by Andrew Randrianasulu 09 Nov '22

09 Nov '22

Because I tend to track this issue https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19232 see latest notes there (you need llvm 15.0.4 - yeah, in this case 0.0.1 version matters) and mesa git + some merge requests manually merged, but results relatively promising - darktable starts and enjoy some 2x speedup (on Power CPU, no less!), Lux render also starts and works. Ffmpeg I guess not tested yet :) for me ffmpeg -filters | grep opencl shows ... avgblur_opencl V->V Apply average blur filter ... boxblur_opencl V->V Apply boxblur filter to input video ... colorkey_opencl V->V Turns a certain color into transparency. Operates on RGB colors. ... convolution_opencl V->V Apply convolution mask to input video ... deshake_opencl V->V Feature-point based video stabilization filter ... dilation_opencl V->V Apply dilation effect ... erosion_opencl V->V Apply erosion effect ... nlmeans_opencl V->V Non-local means denoiser through OpenCL ... overlay_opencl VV->V Overlay one video on top of another ... pad_opencl V->V Pad the input video. ... prewitt_opencl V->V Apply prewitt operator ... program_opencl |->V Filter video using an OpenCL program ... roberts_opencl V->V Apply roberts operator ... sobel_opencl V->V Apply sobel operator ... tonemap_opencl V->V Perform HDR to SDR conversion with tonemapping. ... transpose_opencl V->V Transpose input video ... unsharp_opencl V->V Apply unsharp mask to input video ... xfade_opencl VV->V Cross fade one video with another video. ... openclsrc |->V Generate video using an OpenCL program for tonemapping on OpenCL this line was suggested (on Nvidia) - see below. https://superuser.com/questions/1549479/ffmpeg-4k-hdr-to-1080p-sdr-hardware… so I guess we can use it as example for trying on Radeons (comment about OpenCL 1.2 needed still apply, but rusticl exposes 3.0 :-) ) there is also AUR package, but not sure if it tracks right branch from Karol Herbst (developer) repo .. https://aur.archlinux.org/packages/mesa-rusticl-git I personally look at https://gitlab.freedesktop.org/karolherbst/mesa/-/commits/rusticl/wip_nv/ note enabling this also require some environment flags to be set. ffmpeg -vsync 0 -hwaccel cuda \ -init_hw_device opencl=ocl \ -filter_hw_device ocl -extra_hw_frames 3 \ -threads 16 -c:v hevc_cuvid \ -resize 1920x1080 -i INPUT \ -vf "format=p010,hwupload,tonemap_opencl=tonemap=mobius:param=0.01:desat=0:r=tv:p=bt709:t=bt709:m=bt709:format=nv12,hwdownload,format=nv12" \ -c:a copy -c:s copy \ -c:v libx264 -max_muxing_queue_size 9999 OUTPUT

1 0

shared build fix for my Slackware
by Andrew Randrianasulu 09 Nov '22

09 Nov '22

..where X11 actually located in /usr/X11R7 so this configure works LDFLAGS=-L/usr/X11R7/lib setarch i686 ./configure --with-single-user --without-thirdparty --disable-libdpx --without-libdpx --without-lv2 --disable-static-build

3 9