I can only endorse the words of Andre, great work by you. Such work is also important that the performance stays good. Thanks again. Sam On 26.08.19 18:19, Andre Gronwald wrote:

Thx a lot for your hard work and don't worry about other "missing" things. Everything you deliver makes that software better, and it is already good. And in my country is holiday time, maybe in yours too. And holiday is necessary for everyone. ;-)

Thanks once again to you and goodguy! Regards Andre

Phyllis Smith <[email protected] <mailto:[email protected]>> schrieb am Mo., 26. Aug. 2019, 17:47:

Hi, Today's checkin contains an upgrade from FFmpeg 4.1.4 to 4.2.  No problems are expected in this area and probably later today the patch to use dav1d will also be checked in, but are still working out the logistics and might have some questions.

The 3 week delay since the previous checkin was due to many, many problems coming up with a method and strategy to speed up Feathering (mostly)in Masking for both OpenGL and X11 software driver variations resulting in the same looking output.  Any additional minor or major testing would be appreciated!

Ubuntu16 test build will be available later today.

I did not know that this Masking change was going to take nearly this long, so any commitments to address other issues that I made are no longer going to happen as I had planned.  Oh well.   Phyllis/GG -- Cin mailing list [email protected] <mailto:[email protected]> https://lists.cinelerra-gg.org/mailman/listinfo/cin

Using the mask leads to the following errors at the terminal. The mask is drawn but the fade does not work: it shows the fade all over the frame and not in the mask. The errors, repeated for each action on the mashera tool, are: Playback3D::print_error: link Fragment info ------------- (0) : error C6020: Constant register limit exceeded at n; more than 1024 registers needed to compile program (0) : error C6020: Constant register limit exceeded at dxy; more than 1024 registers needed to compile program (0) : error C6020: Constant register limit exceeded at twh; more than 1024 registers needed to compile program See the video: https://streamable.com/jxw8v

Andrea, gg checked in a change that we have no way to test because I can not generate the C6020 error register limit. When you have time, could you please test? Cinelerra should compile now because he also checked in a change for the Arch meson problem. Let me know. Phyllis (0) : error C6020: Constant register limit exceeded at n; more than

1024 registers needed to compile program

Andrea, OK gg looked at your file and the ffmpeg compile is upset. Please send this log so he can look at that (or you can look at it and see if you see anything wrong): {cin_path}/thirdparty/ffmpeg-4.2/ffbuild/config.log On Tue, Aug 27, 2019 at 3:52 PM Andrea paz <[email protected]> wrote:

Sorry for later. My compile fails again. Attached cin5.log Is better to compile from root? -- Cin mailing list [email protected] https://lists.cinelerra-gg.org/mailman/listinfo/cin

It works! Thank you. If interested, I attach the config.log of when it gave error, and that of the compile went well. However I have no CUDA in the system, I removed it when I realized that my graphics card was too old. If necessary, I can reinstall it and run some tests. But I don't know how useful it is in my system. Also solved the problem with the mask tool. Great! You are the best.

Andrea, gg has looked at your Valgrind output and is happy to report "it is a Perfect run!" The terminal.txt file points some problem with the Undo stack and mask Rotate that he will try to look at today yet. Thank you, GG/Phyllis On Wed, Aug 28, 2019 at 6:10 AM Andrea paz <[email protected]> wrote:

I tried to make a Valgrind, playing a little with the mask tool. I attach the file and also another file where I reported the messages of the terminal that appeared when I used the fade, the rotation, etc. -- Cin mailing list [email protected] https://lists.cinelerra-gg.org/mailman/listinfo/cin

Andrea:

It works! Thank you.

No, thank YOU because it saved us a lot of work before this month's builds. However I have no CUDA in the system, I removed it when I realized

that my graphics card was too old. If necessary, I can reinstall it and run some tests.

Not necessary, good to know about removing it so I keep that in mind.

Also solved the problem with the mask tool.

Thanks for that feedback as I suspect there are other systems out there that would have had the same problem if you had not caught it first. gg/Phyllis

Andrew, Hm, strange - both in cinelerra-goodguy-20190808 and

cinelerra-goodguy-20190827 I can't see Mask's work, if I just add pre-defined shape and play with 'fade' shader. (tried various values between -100 and 100) I tried to add effect (color invesion, brightness ..) but they apply to full track anyway.... Setting track's mixing mode at bay also changes nothing

I also tried X11 and X11-OpenGL output, and 'disable OpenGL' checkbox in mask control window... Project settings were defaults (as after loading single video file), but changing project's ColorFormat from RGBA to RGBA-float, or YUVA changed nothing ...

None of the above is explainable. We do not know how to diagnose the problems you see. Not to be insulting, but did you add an empty video track below the main video track and add the Gradient plugin to that track? That makes it so you can see the mask area. So if you look at the following demo, can you point out where your masking might be going wrong? https://streamable.com/yujpu

It seems online manual a bit outdated in this regard (masking), and giant issue about Masking improvements on bugtracker also disappeared?

Yes, the current manual Masking is totally wrong and Andrea has incorporated the new stuff into the Latex version which is not yet available. The below has the new Masking documentation temporarily: https://www.cinelerra-gg.org/download/Mask_changes.pdf The Bugtracker is still there as issue #2, but it is Closed so that is why you might not see it. You probably have "Hide Status" set to "Closed". But at the top right hand corner of the Bugtracker, you can type in the number "2" for the issue to see it all.