Confirmed that downloading a pristine dev-channel image (build 7978.18.0 for samus) from go/chromoes-images gives the same result.

tsergeant@ or raymes@: do you know who would be the right people to look at this? Seems pretty serious. FWIW, I don't think I've been seeing this on my self-built M51 Chromium OS images.

Did this just start happening recently? Or has it happened for a long time?

In my experience, this has been happening on Samus since R50 was introduced to the Dev channel.

I'm seeing the same with R50 on Guado.

My recollection agrees with #4, though I haven't confirmed for sure (I'll try that now): I believe this is a regression for M50 (hence, I labeled this Type:Bug-Regression).

Confirmed that the same PDF views/scrolls just fine in M49 stable (49.0.2623.95; 7834.60.0 (Official Build) dev-channel samus test). And as mentioned, no problems on my self-built M51 (51.0.2679; 8069.0.2016_03_15_1409 developer-build samus). So it's a M50-only regression I guess?

Thanks all - does it happen reliably? Would a bisect help us track it down?

Happens extremely reliably. I can see the tearing in just a few seconds of navigating around.

What sort of bisect would you like? Just go through historic canary images, for instance? Kernel only? Chrome-only? (I've had trouble self-building Chrome reliably myself, though I've gotten a few builds working. So it'd help if there were pre-builts I could test.)

Or I've got plenty of Samus devices here, if you want me to ship one :)

Bisecting using the canary images would be great. I'm not sure how QA does bisects, +ananthak who might be able to help.

It's strange that it happens on 50 but not 49 or 51. It makes me suspect it may just be a race condition being tickled by something. A bisect would still help to confirm that though.

As pointed out with bug https://bugs.chromium.org/p/chromium/issues/detail?id=596299, this doesn't occur if you set the Pixel to its native resolution of 2560x1700.

Judging by  issue 596299 , then maybe this is more of a graphics issue, and not specific to the PDF reader. And given the only non-samus report is on Guado (also Intel/Broadwell?) maybe this is a kernel/graphics issue. Marcheu?

Anyway, I did a quick bisect. Not finished, but I'm heading home soon. Results:

GOOD: 7837.0.0 (Official Build) dev-channel samus test
GOOD: 7912.0.0 (Official Build) dev-channel samus test
BAD: 7944.0.0 (Official Build) dev-channel samus test
BAD: 7980.0.0 (Official Build) dev-channel samus test

I can follow up more tomorrow, but this is a start.

Cool, thanks! Another data point is that we don't have any significant changes in the pdf rendering code. So I think it would more likely be graphics issue as well.

Final results:

GOOD: 7837.0.0 (Official Build) dev-channel samus test
GOOD: 7912.0.0 (Official Build) dev-channel samus test
GOOD: 7927.0.0 (Official Build) dev-channel samus test
BAD: 7930.0.0 (Official Build) dev-channel samus test
BAD: 7934.0.0 (Official Build) dev-channel samus test
BAD: 7944.0.0 (Official Build) dev-channel samus test
BAD: 7980.0.0 (Official Build) dev-channel samus test

I don't see any more canaries to test.

Browsing the manifeset diff between 7927 and 7930 doesn't show anything all that incriminating, though I wouldn't call myself an expert on most of that...

Any tips on what to do next?

What is the list of changes between 7927.0.0 and 7930.0.0 ?

answering my own question, https://crosland.corp.google.com/log/7927.0.0..7930.0.0

Hmm we don't see chrome rolls in there, do we? I suspect it's a chrome bug, but if we don't see the rolls we will probably have to look for it

It seems like on feb 15 chrome rolled from 50.0.2639.0 to 50.0.2651.0, could it be a problem in there?

Could be...

$ git diff --stat 50.0.2639.0 50.0.2651.0 | tail -1
 12676 files changed, 268016 insertions(+), 230856 deletions(-)

I'm not sure how to study that, nor do I think I'm the right person to do this.

Hi Albert, probably makes sense for someone from UI to drive the investigation here. Can you help find an owner? Thanks!

Ping, Albert?

Moving old issues out of Internal>Graphics to delete this obsolete component ( crbug.com/685425  for details)