visit-dav / visit

VisIt - Visualization and Data Analysis for Mesh-based Scientific Data
https://visit.llnl.gov
BSD 3-Clause "New" or "Revised" License
440 stars 116 forks source link

Lineout cue drawn inches away from mouse pointer gesture path #19544

Open markcmiller86 opened 6 months ago

markcmiller86 commented 6 months ago

Describe the bug

A user on LLNL RZ showed me this was happening running VisIt 3.4.1 in VNCViewer from DELL Laptop. He would drag a lineout and the visual cue would be miles away from the mouse pointer path on the screen. After releasing, it would compute the correct lineout curve and when he returned to the window showing the lineout cue, the cue would be in the correct place.

So, the issue is with the logic that draws and updates that lineout cue during the drag and release event.

He said the badness of the offset varies based on VNC window resolution.

I had Eric try on RZ from his laptop over VNC and everything worked fine. We tried varying VNC resolution with xrandr -s commands of various sorts and the lineout worked fine every time. We did notice that the lineout update drawing was very slow and took a while to catch up to Eric's mouse pointer gestures but it ultimately always settled on the correct cue position.

I also tried 3.4.1 on SCF with VNC viewer and it worked for me as it did for Eric.

cyrush commented 4 months ago

This might be due to using some fancy X11 X-or-ing overlay to do these types of overlays. We should see if there is a simpler way not that we live in "the future".

biagas commented 4 months ago

Might need to look at the vtk mapper classes in src/avt/QtVisWindow/vtk

markcmiller86 commented 4 months ago

The user reporting this is M. Adler

cyrush commented 1 month ago

@brugger1 could this issue be similar to the rubber band zoom issues?

brugger1 commented 1 month ago

The code for rendering the rubber band line looks similar to the zoom lines. It isn't trying to do any fancy XORing. I have tried to reproduce this on my Dell laptop and haven't been able to. Mark wasn't able to reproduce on the closed in a VNC viewer. I believe Mark saw it on the users system so it is real, but I don't know what's special about his system.

Maybe we try on his system with the beta to see if it is still doing it. If it is, maybe we build a version with debugging information that outputs the coordinates of the lines being rendered to see if those are correct.