search

mate-desktop / marco

MATE default window manager
https://mate-desktop.org
GNU General Public License v2.0
197 stars 87 forks source link

panel tray corruption with Marco 1.13 only #247

Closed lukefromdc closed 8 years ago

lukefromdc commented 8 years ago

https://github.com/mate-desktop/mate-panel/issues/394 turned out to be specific to running Marco 1.13 (EDIT: marco 1.12 also affected )on a slow machine (Intel Atom netbook). Sometimes the panel tray background and icon images are corrupted, interacting with them can stop the corruption but changing workspaces corrupts the tray anew. This happens with or without compositing. Does not happen with Marco 1.12 nor with compiz, thus is a new Marco issue.

Since this does not happen with any git master build of marco prior to the 1.13 transition, I suspect some conditional is changing the build on an odd-numbered (dev) version and creating this issue. Similar conditional had to be fixed to let mate-terminal 1.13 build at all.

lukefromdc commented 8 years ago

On further work I found that the last versions of Marco 1.12 also had this problem, but I have to move the tray to invoke it (issue does not always appear with any version so hard to diagnose). More tests coming to get exact first bad version

lukefromdc commented 8 years ago

The last version of Marco 1.10 I have (9-24-2015) is good, so far free of this issue. First version of 1.12 I have is bad-but only with recent versions of mate-panel, so it is some kind of interaction and on fast machines does not occur. I think i still have the 1.11 versions elsewhere, will test them when I can get to them. I've got a debian package built for each commit but not sure if I still have the old 1.11 versions saved anywhere. At any rate this issue was not invoked until recent mate-panel changes, LOTS of background fixes there

lukefromdc commented 8 years ago

OK,I just tested the 1.11 versions I saved as Debian packages. The version I built on Oct 25, whose last commit was https://github.com/mate-desktop/marco/commit/eceb8d7994859c4ebbee493539a6ce6fd2842e50 window: initialize custom_frame_extents

was good. The next version I has was an Oct 31 build of 1.11.1. The only actual code changes there (as opposed to NEWS updates) was the combined merge of the "extents" branch from which the previous good commit and those that led to it came and the "window-move crash fix" branch. One way of invoking the panel corruption is in fact to move the tray on the panel, which is moving a window.

The only commit still shown on the timeline after those branches were merged between the 1.11.1 release and the last good one is https://github.com/mate-desktop/marco/commit/0303e70565fe5a0ede31c2c4591028ed34cbb3a2

window: initialize fullscreen to FALSE

Not sure if that had anything to do with the window move bugfix referred to in my changelogs and in the NEWS file

monsta commented 8 years ago

Oh, https://github.com/mate-desktop/marco/commit/0303e70565fe5a0ede31c2c4591028ed34cbb3a2 actually does nothing. Back then I completely overlooked the fact that the same initialization already was there 10 lines above. Later I removed this excessive initialization in https://github.com/mate-desktop/marco/commit/a04f17f520ffb8d0282dcc1e3a7094b29edc0471.

lukefromdc commented 8 years ago

This now appears to be an Intel driver bug though I am not sure. All mate-panel versions all the way back to 1.10.0 (oldest I have around) affected, marco versions back to well before it showed up also affected. Intel x driver updated Jan 2 and the but showed up only recently and the laptop has not had much use since then.

Also, setting an image background seems to prevent tray corruption entirely, though more testing is needed to be sure. Thus, this appears to be a driver bug hit by fairly recent versions of marco with any version of mate-panel, with a workaround.

If the driver bug still exists when distros ship MATE 1.14 with gtk3, they might need to include in in "known issues." a note to user of pre-Poulsbo Intel Atom netbooks set a panel BG image.

On 1/12/2016 at 6:22 AM, "monsta" notifications@github.com wrote:

Oh, https://github.com/mate- desktop/marco/commit/0303e70565fe5a0ede31c2c4591028ed34cbb3a2 actually does nothing. Back then I completely overlooked the fact that the same initialization already was there 10 lines above. Later I removed this excessive initialization in https://github.com/mate- desktop/marco/commit/a04f17f520ffb8d0282dcc1e3a7094b29edc0471.

Reply to this email directly or view it on GitHub: https://github.com/mate-desktop/marco/issues/247#issuecomment- 170880722

monsta commented 8 years ago

Ok... is this still an issue? Given the previous comments, I really doubt it's Marco misbehaving.

lukefromdc commented 8 years ago

I haven't seen this in many months on my netbook, must have been a driver issue that came and went, invoking some corner case. As far as I am concerned this can be closed

On 3/30/2016 at 3:29 PM, "monsta" notifications@github.com wrote:

Ok... is this still an issue? Given the previous comments, I really doubt it's Marco misbehaving.

You are receiving this because you authored the thread. Reply to this email directly or view it on GitHub: https://github.com/mate-desktop/marco/issues/247#issuecomment- 203596249

monsta commented 8 years ago

Ok, closing it then.