951433 - [B2G][Calendar] Fast horizontal swipes while in Week view tab results in partialy-blank page

Status: RESOLVED DUPLICATE of bug 966507

(Firefox OS Graveyard :: Gaia::Calendar, defect)

Description

[Nikolai Khristoforov]

Attachment: Logcat

While in the Week tab, if you perform a fast swipe in a horizontal direction to change the viewable days, the page becomes blank under the All Day row. This does not reproduce with every swipe, but once you get it to reproduce once it becomes easier.

Repro Steps:
1) Updated Buri to BuildID: 20131216140918
2) In Settings > Device Information > More Information > Developer, turn on "Enable APZ for all content," "Remote debugging," and "Gaia debug traces."
3) Press the Home button and open the Calendar app.
4) Select the Week tab on the bottom of the page.
5) Quickly swipe horizontally to change the viewable days. This step will probably need to be done multiple times to reproduce the bug.

Actual:
The page becomes blank under the All Day row.

Expected:
The viewable days would change without any part of the page becoming blank.

Environmental Variables:
Device: Buri 1.3 MOZ
BuildID: 20131216140918
Gaia: 504cf9a988cd84760b4040706ccc41e508a97fd2
Gecko: 55b75ce2263e7d2a8cbfdabc7e5206d9caa89f98
Version: 28.0a2
Firmware Version: V1.2_20131115

Repro frequency: 7/10
See attached: Logcat and Screenshot

Comment 1

[Nikolai Khristoforov]

Attachment: Screenshot

Comment 2

[Nikolai Khristoforov]

This bug does not reproduce when the "Enable APZ for all content" option is not selected.

Comment 3

[Vivien Nicolas (:vingtetun) (:21) - (NOT reading bugmails, needinfo? please)]

I can not reproduce that. So I would bet that it is a duplicate of bug 947337. I have a patch I run locally for this bug. I will try to remove it from my tree to see if I can reproduce without it.

Comment 4

[Vivien Nicolas (:vingtetun) (:21) - (NOT reading bugmails, needinfo? please)]

Nikolai, can you produce a video ? I tried very hard but I can not reproduce (even without the patch I mentioned).

I would like to see how you are swiping exactly to see if I'm missing something. Also do you use a special set of calendar data, or do you use the reference workload (ligh/medium/heavy?).

thanks.

Comment 5

[Nikolai Khristoforov]

I've uploaded the video here: http://youtu.be/FVCME2IxBTI

I noticed that it was easier for me to reproduce when I was in the middle of the calendar instead of at the top, so you can try scrolling down first like I did in the video. As for the calendar data, I did not use a special set so I think it was the reference workload.

Comment 6

[Vivien Nicolas (:vingtetun) (:21) - (NOT reading bugmails, needinfo? please)]

(In reply to Nikolai Khristoforov from comment #5)
> I've uploaded the video here: http://youtu.be/FVCME2IxBTI
> 
> I noticed that it was easier for me to reproduce when I was in the middle of
> the calendar instead of at the top, so you can try scrolling down first like
> I did in the video. As for the calendar data, I did not use a special set so
> I think it was the reference workload.

Thanks for the video.

Comment 7

[Vivien Nicolas (:vingtetun) (:21) - (NOT reading bugmails, needinfo? please)]

(In reply to Vivien Nicolas (:vingtetun) (:21) from comment #6)
> (In reply to Nikolai Khristoforov from comment #5)
> > I've uploaded the video here: http://youtu.be/FVCME2IxBTI
> > 
> > I noticed that it was easier for me to reproduce when I was in the middle of
> > the calendar instead of at the top, so you can try scrolling down first like
> > I did in the video. As for the calendar data, I did not use a special set so
> > I think it was the reference workload.
> 
> Thanks for the video.

I have an hard time to reproduce. I see sometimes the white area but then it quickly repaints and everything is fine after that.

Comment 8

[Botond Ballo [:botond]]

I can reproduce this occasionally, if I swipe back-and-forth quickly, instead of just in one direction.

The white area always goes away the next time I touch the page.

Comment 9

[Botond Ballo [:botond]]

(In reply to Botond Ballo [:botond] from comment #8)
> I can reproduce this occasionally, if I swipe back-and-forth quickly,
> instead of just in one direction.
> 
> The white area always goes away the next time I touch the page.

Assuming that what Vivien and I can repro is all there is to this bug, based on the low impact on user experience (the white screen only appears occasionally, after an uncommon gesture (swiping back-and-forth quickly), and going away the next time you touch the screen) I would suggest not blocking on this bug and moving it to be a dependency of bug 949585 (follow-up gaia-apzc issues) instead. If there are no objections I will do that.

Comment 10

[Milan Sreckovic [:milan] (needinfo for best results)]

Right now, this is 1.3? - we'll cover this in the triage, and if it gets decided that it is 1.3+, we can move it back to block bug 909877.  For now, moving to block bug 949585 which are the apzc follow up bugs.

Comment 11

[Miller Medeiros [:millermedeiros]]

I still don't have a device to try it out (so I can't replicate the bug) but I guess one solution would be to do some sort of "throttle/debounce", to make sure the rendering of the new calendar only happens after a few milliseconds. That way swiping 10x in a row would only trigger a single repaint. Maybe that would be enough to fix the bug and would also reduce the amount of unnecessary work by the CPU (better performance, better battery life). - this change would probably make it easier to add some sort of transition in the future.

I can provide a patch for you guys to try it out, just not sure on how to make the integration test for this case - since it wouldn't catch the bug I think it's kinda pointless to update the tests (current test already checks if date is changed during swipe, so it looks good enough for me).

Comment 12

[Jason Smith [:jsmith]]

Talking with Naoki offline, this is hard to reproduce & will be tough for a user to trigger. On that regard, we won't block on this.

Comment 13

[Kartikaya Gupta (email:[e-mail])]

Based on the STR and symptoms described here I think that my patches in bug 951113 might fix this issue. Once that lands it would good if QA could retest this to see if it still happens.

Comment 14

[Jason Smith [:jsmith]]

(In reply to Kartikaya Gupta (email:[e-mail]) from comment #13)
> Based on the STR and symptoms described here I think that my patches in bug
> 951113 might fix this issue. Once that lands it would good if QA could
> retest this to see if it still happens.

Feel free to flag qawanted when that patch lands.

Comment 15

[Kartikaya Gupta (email:[e-mail])]

Bug 951113 is on central now; requesting qawanted to retest this bug.

Comment 16

[Tony Nguyen]

Attachment: screenshot

This issue is still present on today's Buri 1.3 build and I'm attaching a screenshot to represent it.

Environmental Variables:
Device: Buri v1.3 mozRIL
BuildID: 20140122004001
Gaia: 744fb691c2b2a25a07c5d19fabf5748ae9aba4d9
Gecko: 71a8786c3815
Version: 28.0a2
v1.2-device.cfg

Comment 17

[Botond Ballo [:botond]]

Unassigning for now as my plate is full with 1.3+ blockers and I'm unlikely to get to looking at this in the next couple of weeks. Please feel free to reassign later.

Comment 18

[Kartikaya Gupta (email:[e-mail])]

This looks like another dupe of bug 966507. With the patch from that bug applied onto latest 1.3 code I can no longer reproduce this problem, whereas it's pretty easy for me reproduce without that patch.