Mouse hits invisible walls in conquest

[msharov]

Shareware game conquest: https://drive.google.com/file/d/101UOmEygho5MU1W9Ov4JdUMQwGNZU2uG
Played in DOSBox, fullscreen, using i3wm. Worked without problems with sdl2.

While playing the game the mouse occasionally hits a wall. That is, the cursor hits a line, either horizontal or vertical, at some apparently random location on the screen, and moving the mouse further does not move the cursor past the line. Moving the mouse sideways can slide the cursor along the line. Moving the mouse in the opposite direction usually resets the problem and the wall disappears. All this happens randomly; with no apparent pattern.

The mouse also feels more sensitive, moving too fast for comfort, although it is not obvious whether the actual speed has changed or just become more jerky. This may or may not be related to the walls.

[slouken]

@Kontrabant, can you take a look at this one?

[Kontrabant]

I tried this in i3 on both sdl2-compat 2.32.56 and the 2.32 branch of real SDL2 and couldn't notice any tangible difference between the two, and never hit the 'wall' issue in a few minutes of moving the mouse around. I tried a few other desktops as well with X/XWayland/Wayland, and didn't notice anything abnormal on those either. Note that the initial report didn't mention what variant of dosbox they were using, so I used dosbox-staging, as that seems to be what most distros ship in their repositories now.

Given that this report is several months old, the question for @msharov is if this is still happening with the current sdl2-compat + SDL3 releases? Was this possibly fixed by the mouse coordinate accumulation changes? A code search of dosbox-staging and dosbox-x shows that they don't grab, warp, or otherwise restrict the pointer in any way, so this is/was probably coordinate related.

[msharov]

Unfortunately, it still happens with the sdl2 2.32.56-2, sdl3 3.2.22-1, dosbox 0.74.3-3, i3-wm 4.24-1. This on Arch, standard packages. I tried it twice just now, with the mouse hitting a wall during the first game round on the first try and during the second round on the second try. In both cases the cursor hit a vertical wall, the first time it was at about 1/3 of screen X, the second time maybe an inch off the monitor edge. Moving the mouse to the right edge resets it. The first time I touched the right edge briefly, I think, and the "wall" moved to the left a few inches, as if the screen scrolled a virtual desktop. On the second attempt the wall was already too close to the edge, so I couldn't do it again. I tried a few more times, and was eventually able to hit an upper right corner again. The virtual desktop scrolling thing happened again, I was able to bump up the corner several times by briefly touching the bottom edge. So still a hit and miss on reproduction, but I can make it happen often enough.

[Kontrabant]

I built the original dosbox, but still can't seem to hit the invisible wall issue on i3.

Would you be able to test with a newer version of dosbox, such as dosbox-staging? dosbox 0.74 is very old and still uses SDL 1.x, so there are two compatibility layers in the mix (sdl12-compat→sdl2-compat→SDL3). This would help narrow down if it is some kind of bug in the interaction between sdl12-compat and sdl2-compat.

[msharov]

Sorry, building dosbox-staging requires many additional AUR packages, and will likely take a very long time to figure out. Feel free to close the bug as unreproduceable, I guess, since I just don't have enough motivation to put in this much effort. I'll blacklist std2-compat permanently, which keeps things working for me, and you can debug more if someone finds a more reliable reproduction.