Bug 113664 - Cell selection rectangle is unevenly thick based on vertical position (GPU-dependent, happens on AMD)
Summary: Cell selection rectangle is unevenly thick based on vertical position (GPU-de...
Status: RESOLVED WORKSFORME
Alias: None
Product: LibreOffice
Classification: Unclassified
Component: graphics stack (show other bugs)
Version:
(earliest affected)
5.3.0.3 release
Hardware: All Windows (All)
: medium normal
Assignee: Not Assigned
URL:
Whiteboard:
Keywords: bibisected, bisected, regression
Depends on:
Blocks: VCL-OpenGL
  Show dependency treegraph
 
Reported: 2017-11-05 20:42 UTC by Aron Budea
Modified: 2019-04-17 19:25 UTC (History)
4 users (show)

See Also:
Crash report or crash signature:


Attachments
Screenshot (52.51 KB, image/png)
2017-11-05 20:42 UTC, Aron Budea
Details
Bibisect log (2.90 KB, text/plain)
2017-11-06 14:56 UTC, Telesto
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Aron Budea 2017-11-05 20:42:51 UTC
Created attachment 137542 [details]
Screenshot

When OpenGL is enabled, the thickness of the selection rectangle depends on the cell position.
See the attached screenshot. Likely Windows-only.

Repro with LO 6.0 master build (d8f0e4144d51ff5709d005d289d71aa32f323913), 5.3.0.3 / Windows 7.
Not reproduced with LO 5.2.0.4.
=> regression

Early 5.3 builds with OpenGL enabled are very shaky on my system, all I can say it's in the first ~1k commits of 5.3 based on bibisect-win32-5.3.
Comment 1 Telesto 2017-11-06 09:17:35 UTC
No repro
Version: 6.0.0.0.alpha1+
Build ID: b3f1d199a72ce87cb65ddaeac922564f57da6a4d
CPU threads: 4; OS: Windows 6.3; UI render: GL; 
TinderBox: Win-x86@42, Branch:master, Time: 2017-11-06_00:10:53
Locale: nl-NL (nl_NL); Calc: CL
Comment 2 Aron Budea 2017-11-06 10:29:25 UTC
To be more precise, it depends on vertical position. If I start from row 1 and move down, it alternates between top and bottom being thick, somewhat thinner and thin, sometimes they match (eg. thin - thin), but not always.

Ok, this might be GPU-dependent, I just tried with an Intel Ivy Bridge-series IGP (3xxx series CPU), and the thickness was even there (had to force OpenGL, though).
The system where it happens has an AMD RX 570.
Comment 3 Telesto 2017-11-06 11:23:45 UTC
My bad, I can reproduce this
Version: 6.0.0.0.alpha1+
Build ID: b3f1d199a72ce87cb65ddaeac922564f57da6a4d
CPU threads: 4; OS: Windows 6.3; UI render: GL; 
TinderBox: Win-x86@42, Branch:master, Time: 2017-11-06_00:10:53
Locale: nl-NL (nl_NL); Calc: CL
Comment 4 Telesto 2017-11-06 11:28:22 UTC
Is this zoom-level depended?
Comment 5 Aron Budea 2017-11-06 12:16:49 UTC
(In reply to Telesto from comment #4)
> Is this zoom-level depended?

Yes, it shows different thickness for the same cell when I zoom in/out.
Comment 6 Telesto 2017-11-06 14:56:59 UTC
Created attachment 137574 [details]
Bibisect log

Regression introduced by:

author	Tomaž Vajngerl <tomaz.vajngerl@collabora.co.uk>	2016-05-29 06:36:39 (GMT)
committer	Tomaž Vajngerl <tomaz.vajngerl@collabora.co.uk>	2016-06-08 02:39:22 (GMT)
commit 2003076c4318511a3d621558d3b44b4e8e6c6529 (patch)
tree ecc2742bb911d2960ecd96f320a802ff8d43816f
parent 344dc6939c45552dc162ea8b3f892e2ae4998d64 (diff)
opengl: batch draw polypolygons
Comment 7 Aron Budea 2017-11-11 12:23:47 UTC
Thanks for the bibisect! Let's CC Tomaz Vajngerl.
Comment 8 Xisco Faulí 2018-06-05 19:54:00 UTC
Adding Cc: to Tomaž Vajngerl
Comment 9 Miklos Vajna 2019-04-17 15:04:48 UTC
Does this still happen? I ask as bug 107966 was bisected to this commit as well, and that was fixed last month.

I can't reproduce this problem on Windows with my intel and nvidia cards, though -- even if I revert the fix.
Comment 10 Aron Budea 2019-04-17 19:25:39 UTC
Interestingly I can't repro it with a current master build, or even with the same state as the original repro. I wonder if the GPU driver got changed... anyway, let's close as WFM.