Open ghost opened 12 years ago
Hi Plamen,
Perhaps the randr commit I pulled from you is no longer needed with the code rewrite? I've reverted it in branch rotate_revert (6e5a9b64e9) can you try that one?
Kind regards, Tias
Hello Tias,
I found some time to test the revert but unfortunately it doesn't give a different result.
Best regards, Plamen
That's unfortunate. I wouldn't know what else can cause this behaviour.
Can you git-bisect to find the offending patch?
Thanks, Tias
Hello Tias,
Sorry for not replying for such a long time.
I bisected the master branch and found out that commit 07978b90b81237d83071426643416161087e45dd breaks the calibration when the screen is rotated. I tested a fix for this exact issue in my test branch (299b69de8e9fc3fce5df56a906fca8afab49a3e4).
Later I took a look at Forest and Jeff's pull requests (#36 and #45). I segfaulted on Forest's build, but didn't check why, so I don't know if it fixes this issue. Jeff's build did run though and it does fix this problem. Funnily enough it also fixes my other problem where axes are swapped on rotated screen on the first calibration.
Best regards, Plamen
Hi,
@Tokoyami do you mind building with my changes and running under gdb to get a stack trace?
Thanks, Forest
Hello Forest,
I actually already found the problem. I'll post it in a bit in your pull request.
Best regards, Plamen
Hi Plamen,
Thanks, that would be great.
-Forest
On Tue, Feb 05, 2013 at 11:24:55AM -0800, Plamen Kolev wrote:
Hello Forest,
I actually already found the problem. I'll post it in a bit in your pull request.
Best regards, Plamen
— Reply to this email directly or view it on GitHub.
*
I've tested the current master (9b0803bf0a) on our two Elo screens and normal/inverted rotation calibration works fine. The problem comes when we rotate the screen to the left or right. Then the returned calibration output looks like this: https://gist.github.com/3031786. Compared to the state before the changes from 21/6 onward, it's absolutely broken. Before our issue was that when rotated left/right the axis invertion would be wrong and a second run would fix it. Now it's unusable.