BQ: Thin column of pixels visible on the left (wrapped from the right)

[Oliver Grawert]

with image 81 my krillin shows a white 1px line on the left edge of the screen, specifically visible when switching apps or sliding the launcher in.

[Albert Astals Cid (aacid)]

Does it fix itself by locking/unlocking the screen a few times? I think i saw that in the past and then went away and now maybe is back.

I am not totally sure that's unity8 fault tbh.

[Daniel van Vugt (vanvugt)]

Mir's been unchanged for a while now - still on version 0.7.3 in image 84 and I can't see any while line on krillin.

Sounds more likely something has changed in unity8.

[Oliver Grawert (ogra)]

yes, it seems to have vanished during the day after a few lock/unlock cycles.

[Michael Zanetti (mzanetti)]

Daniel, this is not really a new issue. I saw that from the first day I booted Krillin. However, it does recover itself quite easily which is why it gets mostly unnoticed.

[Oliver Grawert (ogra)]

i am pretty sure we have another bug like this from months ago, i have filed a new one because i havent seen it in months (and i cant reproduce it at all now)

[Andrea Cimitan (cimi)]

Cannot reproduce now... can we provide a use case? steps?

[Andrea Cimitan (cimi)]

I have it today... 116, looks like the left column of pixels is not updated and shifted down 1 px. phablet-screenshot does not have the bug

[Michael Zanetti (mzanetti)]

hmm, if the screenshot doesn't show it while the device screen does, I'm having a hard time to believe this is actually caused by unity. Would rather suspect the GPU driver or similar.

[Launchpad Janitor (janitor)]

[Expired for unity8 (Ubuntu) because there has been no activity for 60 days.]

[Sturm Flut (sturmflut)]

This bug gets triggered much more often since krillin r23 (OTA-4), at one point I got it on four out of 18 reboots. It has to be an issue at the driver level, since screenshots taken with the device itself show a correct picture, just the display doesn't.

[Sturm Flut (sturmflut)]

I rebooted my krillin until the bug got triggered (about ten times) and tried to gather as much information as possible directly after the device was accessible via ADB. I've attached the following files:

- dmesg
- Output of "logcat -b main"
- Output of "logcat -b system"
- Contents of all GPU-related directories under /sys/kernel/debug
- Contents of all system- and GUI-related logfiles under /var/log/

I am currently trying to find any differences to a "normal" environment, but no luck so far.?field.comment=I rebooted my krillin until the bug got triggered (about ten times) and tried to gather as much information as possible directly after the device was accessible via ADB. I've attached the following files:

- dmesg
- Output of "logcat -b main"
- Output of "logcat -b system"
- Contents of all GPU-related directories under /sys/kernel/debug
- Contents of all system- and GUI-related logfiles under /var/log/

I am currently trying to find any differences to a "normal" environment, but no luck so far.

[Michael Zanetti (mzanetti)]

Reassigning to canonical-devices-system-image as this is not unity8's fault but I don't know where to put it. Pat, please dispatch it to the correct project. Seems it is related to the display driver.

[John McAleely (john.mcaleely)]

Also tracked in OEM bug #1351554

[Michal Predotka (mpredotka)]

Just to let you know I had this issue after updating my bq e4.5 to OTA-10 today. There was about 1-2px "line" on the left of the screen. It went away after few minutes. I made a screenshot but it wasn't visible on it (as mentioned before in comments).

[Daniel van Vugt (vanvugt)]

Confirmed again by duplicate bug 1586646.

[Daniel van Vugt (vanvugt)]

We have another open bug about the Unity8 screen not fitting the physical screen: bug 1583469

I think they might be related.

[Daniel van Vugt (vanvugt)]

Oh I can finally see this bug on krillin today.

The problem is rather obvious -- we're displaying from the wrong framebuffer address. Under a magnifying glass it looks like we're scanning out from 3 pixels or so before the actual start of the framebuffer. So the whole display is shifted and wrapped a few pixels to the right.

All of this happened randomly after some swipes and taps; without installing any apps and without connecting any peripherals.

After locking and unlocking the phone the bug went away and the screen is now correct again.

[Daniel van Vugt (vanvugt)]

I should add that when the bug occurred I could see about 3 bright (uninitialized) pixels in the top left corner, showing where we underrun the framebuffer.

[Alberto Mardegan (mardy)]

I never saw this bug in about two years of usage of my phone, but today I saw it for the first time, right after the phone rebooted after installing OTA-12. An additional reboot made the issue go away.

[Daniel van Vugt (vanvugt)]

This might be a bug in Mir's Android overlay logic. But I think very unlikely Unity8 is to blame.

[Daniel van Vugt (vanvugt)]

Won't Fix might not be appropriate if this is something like an uninitialized variable affecting the framebuffer offset... But hopefully this is a device-specific bug not in the Mir code.