Issue metadata
Sign in to add a comment
|
Regression: Weird blue border is seen while closing the 'Report An Issue' overlay.
Reported by
lpa...@etouch.net,
Mar 10 2016
|
||||||||||||||||||||||||
Issue descriptionChrome Version: 50.0.2661.26 0abdf9cbb43ea5639687aa21a2bcb206b1dc6d7b-refs/branch-heads/2661@{#169} (32/64 bit) OS: Windows 10, Windows 8 Steps: 1. Launch chrome and navigate to a NTP (new tab page) 2. Go to Wrench >> Help >> Report an issue... 3. Click on the close button and observe the overlay. Actual: Weird blue border is seen while closing the 'ReportAnIssue' overlay. Expected: No such blue border should be seen while closing the 'ReportAnIssue' overlay. This is a regression issue broken in M-50. Manual Regression Range: Good Build: 50.0.2625.0 Bad Build: 50.0.2627.0 Providing Change log Url as chromium builds does not show feedback overlay: https://chromium.googlesource.com/chromium/src/+log/50.0.2625.0..50.0.2627.0?pretty=fuller&n=10000 Suspecting: r370564 ? Please re-assign if your change is not the cause of this issue. Note: Issue is seen only in Windows 10, Windows 8 the same is not reproducible in Windows 7, Linux and Mac OS.
,
Mar 10 2016
,
Mar 10 2016
I don't see a difference between the actual and expected video files, nor do I see anything wrong when trying to reproduce myself. Can you explain in more detail what the bug is?
,
Mar 15 2016
lpanse@, can you please update the thread per c#3. Thank you!
,
Mar 16 2016
With response to comment #3: Weird blue border is seen while closing the 'ReportAnIssue' overlay, no such blue border should be seen. Attached the screenshot for reference. @pdr - Can you please take a look!!
,
Mar 16 2016
Possibly https://chromium.googlesource.com/chromium/src/+/580431e42ab952e84b32e9244acb252b1a76c575? This is really a guess but it's the most likely candidate I see in that range. If it's not that change, I think we will need a proper bisect on windows.
,
Mar 16 2016
+rkc@ who is the owner of the feedback extension. The change cited in #6 affects the closure of menus. Such as the wrench menu seen in the video. However the closing of the feedback window is separate. We likely need a proper bisect on windows.
,
Mar 16 2016
,
Mar 17 2016
It's probably not a Feedback app regression issue. There are no Feedback App CLs in the range https://chromium.googlesource.com/chromium/src/+log/50.0.2625.0..50.0.2627.0?pretty=fuller&n=10000 Could you please triage to someone who can bisect on Windows? I don't have a Windows workstation.
,
Mar 17 2016
I won't have a Windows workstation for a bit. pdr@ do you know someone who could bisect further?
,
Mar 21 2016
Putting this on the paint team radar. We're going to bisect this manually. I'm getting a windows box setup right now and expect to have it up and running by the end of the week.
,
Mar 22 2016
I see no problem on m49, m50 or canary on Windows 7. This might only repro on Window 8 or 10.
,
Mar 22 2016
schenney@, As per the initial bug report "Issue is seen only in Windows 10, Windows 8 the same is not reproducible in Windows 7, Linux and Mac OS" Thank you!
,
Mar 28 2016
I had hoped to get to this last week but didn't get past installing windows 8. Next week isn't looking much better, since chrishtr, wkorman, and I all have high priority issues on our plates. This is likely a real bug so I'm leaving it as a P1 on the paint team so it doesn't get lost. @Test team, is there any way to get a narrower bisect on this bug using the bisect script instead of the official builds?
,
Mar 28 2016
Stephen can you reproduce this?
,
Mar 28 2016
I have no machine on which I can repro it, as the report says it only applies to Win 8 or 10 (comment #13). Maybe we need to pass this to the Windows specific team to verify and track down for us. Otherwise we need to figure out how to get such a machine. Given I'm about to go on vacation for 2 weeks, I don't think I'm a good person to do that. Back to chrishtr@ for triage. There seems to be no Component to attract the attention of people who manage the Windows platform.
,
Mar 28 2016
A friendly reminder that M50 Stable is launching soon! Your bug is labelled as Stable ReleaseBlock, pls make sure to land the fix and get it merged into the release branch by Apr-5. All changes MUST be merged into the release branch by 5pm on Apr-8 to make into the desktop Stable final build cut. Thanks!
,
Mar 28 2016
From the regression range in original report, ignoring Skia rolls and Angle rolls, both of which might matter ... afakhry@, could you look through this list and tag anything that seems likely. The cause of this bug depends a lot on how the feedback overlay is styled and painted. Then we need to find someone who has the necessary dev environment to try reverting patches. https://chromium.googlesource.com/chromium/src/+/a5c571709372144f5175b08a9aaacdd94b8f93d7 is a Win8+ specific change to compositor code. Seems highly likely given the platforms on which we have the bug. https://chromium.googlesource.com/chromium/src/+/38858c57782f2a32cf4f93fcfcc35c340e665be4 actually seems highly likely if the offending blue area is related to a shadow or other filter effect. It could also conceivably be platform specific code. https://chromium.googlesource.com/chromium/src/+/01b20b2ffc26665131edf2df5709a3c226439107 deals with painting outlines, which may be relevant if skipping outline painting somehow causes us to leave something unpainted or paint the wrong thing. Maybe there's an error in the logic for when to skip? https://chromium.googlesource.com/chromium/src/+/537c5c3f3a4f96823e5923694eda37a34add3cbe, is all about display configuration for extensions, so seems a definite candidate. https://chromium.googlesource.com/chromium/src/+/845e85601d268cc9fbccf95a2baf7870166f979a this deals with surface damage notification, which might be relevant if none of the above make a difference.
,
Mar 28 2016
jschuh@ Could you please help assign this to someone who works on Chrome on Windows? Thanks.
,
Mar 28 2016
kylixrd@, bsep@ - Could one of you help diagnose this? Feel free to drag ananta@ into it if needed. P.S. I'm going to assume that everyone on this bug has already heard at least one of my rants about having developers on your teams who regularly build and test on Windows. If you haven't, please stop by for one of my upcoming orientation sessions.
,
Mar 29 2016
There is no option available called "Report an issue" for chromium builds, so unable to provide tool bisect. Narrowed down the bisect manually and providing the regression range below. Good Build:50.0.2626.0 Bad Build: 50.0.2627.0 ========================= Change Log as per omahaproxy: https://chromium.googlesource.com/chromium/src/+log/50.0.2626.0..50.0.2627.0?pretty=fuller&n=10000 From the above change log suspecting below Review URL: https://codereview.chromium.org/1517693002 senorblanco - Could you please check whether this is caused with respect to your change, if not please feel free to remove your name from this bug. Thanks!
,
Mar 30 2016
It's happening even in stable (on closing AND opening the dialog) so I don't think this is actually a regression. (49.0.2623.87 (Official Build) m (32-bit), Windows 8.1) I don't see it on my Windows 10 machine on stable, but I do see it on canary. It's also harder to spot there because the "weird border" is the same white as the dialog box, so it could just be that the performance is good enough on stable you can't see it. Anyway I'm not sure this needs to block stable release.
,
Apr 4 2016
Could anyone review the blocker label and remove it, if this is not blocking.
,
Apr 4 2016
Removing 'RBS' as per c#22, please feel free to update the thread if someone feels otherwise. Thank you!
,
Apr 14 2016
Re-checked the issue in latest canary 52.0.2708.0 and Dev 51.0.2704.7, issue is reproducible in both the builds.
,
Apr 18 2016
@kylixrd: Friendly Ping! Hey, would you mind providing an update on the above issue? Appreciate your response. Thank you!
,
Apr 18 2016
The issue is significantly more visible on Windows 8.x than on Windows 10. From what I can see, when Windows does the fade-in/out animation, the window content (client area) isn't painted and only the frame border (non-client) is visible. This results in that prominent blue border (the default Windows 8.x color scheme).
,
Apr 21 2016
I wonder what the client area insets and DwmExtendFrameIntoClientArea parameters are for this window. It seems like it we want it to have a client area equal to the window rect, so if DWM isn't extending the frame into the client area then I wouldn't expect it to draw the window borders. It's possible that this is related to enabling DirectComposition. It seems that windows has a bug on hiding the window where it'll hide the DirectComposition contents before the fade animation is complete. However I disabled DC on Win8.1 and Win8, so I'm not sure why it's happening there.
,
Apr 26 2016
@kylixrd: Is there any update on the above issue ? Appreciate your help. Thank you!
,
Apr 26 2016
There is a CL under review for this: https://codereview.chromium.org/1903223005/ It's turning into a larger issue than originally thought. jbauman@ is on the right track here. The issue is, in fact, DwmExtendFrameIntoClientArea. However, changing this to 0 also eliminates the activation highlight around the window which tends to cause the window to blend into the windows under it.
,
Apr 28 2016
,
May 3 2016
,
May 6 2016
The weird blue overlay still sen on Win 10 using canary 52.0.2726.0. Launch Chrome >> NTP >> Click on wrench Menu >> Help >> Report an issue >> Observe a blue overlay
,
May 9 2016
@kylixrd: Hey, Would you mind providing an update on the above issue ? Thank you!
,
May 9 2016
No more work has been done on this since it is actually an issue with the the apps API, and not specifically the issue report dialog. It can happen with any Chrome app that doesn't have a frame.
,
Jun 1 2016
Issue is still reproducible on Win 10 , Win 8 using chrome canary 53.0.2754.0.
,
Jun 3 2016
@kylixrd: does this bug need a new owner then? Is the component wrong?
,
Jun 3 2016
,
Jun 3 2016
The blue border issue is an artifact of how Windows handles the DwmExtendFrameIntoClient API along with Chrome's handling of WM_NCCALCSIZE and WM_NCPAINT. Eventually, the goal is to have Chrome paint the whole frame rather than allow Windows to paint some of it. bsep@ was, at one point, working on this very thing.
,
Jun 17 2016
Issue is still reproducible on Win 10 [Pro] for Google Chrome Canary Version - 53.0.2770.0
,
Jun 24 2016
Tested the above issue on Windows-8 OS using Latest Canary chrome version 53.0.2778.0 (Official Build) . It's still reproducible from our end. Attaching screen-cast for the same.
,
Jun 28 2016
Gentle Ping. @@kylixrd: Still able to repro this issue on Chrome Canary Version - 53.0.2782.0
,
Jun 30 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/03e3572cecc4f206b260dbeddf40172d46217617 commit 03e3572cecc4f206b260dbeddf40172d46217617 Author: kylixrd <kylixrd@chromium.org> Date: Thu Jun 30 01:02:59 2016 This will reduce or eliminate the momentary display of the window frame when showing/hiding the window. It does this by adding/removing the frame extending into the client upon showing or hiding using DwmExtendFrameIntoClientArea, respectively. BUG= 593676 Review-Url: https://codereview.chromium.org/1903223005 Cr-Commit-Position: refs/heads/master@{#403046} [modify] https://crrev.com/03e3572cecc4f206b260dbeddf40172d46217617/ui/views/win/hwnd_message_handler.cc [modify] https://crrev.com/03e3572cecc4f206b260dbeddf40172d46217617/ui/views/win/hwnd_message_handler.h
,
Jul 6 2016
,
Jul 7 2016
Verified the issue on Windows 10 using chrome latest M53-53.0.2785.8 by following steps mentioned in the original comment. Observed no blue border is seen while closing the 'Report An Issue' overlay. Hence adding TE-Verified label.
,
Nov 22 2016
|
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by ranjitkan@chromium.org
, Mar 10 2016