New issue
Advanced search Search tips

Issue 907133 link

Starred by 1 user

Issue metadata

Status: Assigned
Owner:
Cc:
EstimatedDays: ----
NextAction: ----
OS: Linux
Pri: 1
Type: Bug

Blocked on: View detail
issue 907554
issue 907557



Sign in to add a comment

leak_detection_linux DCHECK failures

Project Member Reported by huangdarwin@chromium.org, Nov 20

Issue description

The leak_detection_linux bot is failing on 2 websites, apple.com and goal.com. This is likely due to error on the bot and not on CLs, so the bot should be fixed (or removed). Errors are the same on different CLs with different purposes/code touched.

Repro steps:
Create a CL branched from master, with minimal code changes. Some sort of change is necessary, otherwise the bot will simply “succeed” and not run due to no diff from master.

Expected Result:
No failures

What happens instead?
Failure, likely regarding missing machine.

Example failures:
https://ci.chromium.org/p/chromium/builders/luci.chromium.try/leak_detection_linux/25 for https://crrev.com/c/1341145 (small change)

https://ci.chromium.org/p/chromium/builders/luci.chromium.try/leak_detection_linux/24 for https://crrev.com/c/1336440 (larger/different change, but same errors)

All these CLs failed this leak_detection_linux bot, and all on the same type of errors (goal.com and apple.com crash). Errors seem to be about a missing machine.

Notable repeated error between apple.com and goal.com error logs:
Error: /etc/machine-id contains 0 characters (32 were expected).
Failed to connect to the bus: Could not parse server address: Unknown address type

 
Components: -Infra>Platform>Buildbot Infra>Client>Chrome
CCI, if you think this is a bot issue please pass this to labs. Thanks!
Components: -Infra>Client>Chrome
Owner: huangdarwin@chromium.org
Status: Assigned (was: Untriaged)
Summary: leak_detection_linux DCHECK failures (was: leak_detection_linux buildbot failure: likely missing machine)
This doesn't look like an infra issue at all; this looks like a couple of relatively normal DCHECK failures.

-> huangdarwin, if you have any specific reason why you think the machine-id log line you mentioned would cause these DCHECK failures, please elaborate. If you do so, feel free to re-add component:Infra>Client>Chrome & unassign yourself.

from apple.com:

  [1:1:1119/110816.522488:FATAL:layout_flexible_box.cc(1011)] Check failed: content_size >= LayoutUnit() ("-2.1875" vs. "0")

  https://codesearch.chromium.org/chromium/src/third_party/blink/renderer/core/layout/layout_flexible_box.cc?rcl=622c02f9a7e9fe8d55bab94de43b3dab30b8b9b8&l=1011

from goal.com: 

  [1:1:1119/111113.423679:FATAL:paint_chunks_to_cc_layer.cc(440)] Check failed: state_stack_.size(). Error: Chunk has an effect that escapes layer's effect.

  https://codesearch.chromium.org/chromium/src/third_party/blink/renderer/platform/graphics/compositing/paint_chunks_to_cc_layer.cc?rcl=622c02f9a7e9fe8d55bab94de43b3dab30b8b9b8&l=440
Ah fair, thank you for the catch. I seem to have misread the failure logs, sorry. 

I'll set up appropriate bugs so that this DCHECK issue gets fixed and we can use the bot more effectively for leak detection again.
Blockedon: 907554
Blockedon: 907557
Cc: pwnall@chromium.org
Is there anything left in this bug, or were the all issues captured in the newly filed bugs?
All issues that are causing the bot to fail are captured in the newly filed (Blockedon) bugs. As far as I know, if those two DCHECKs being hit are fixed, the bot should be green on master again.
Cc: jbudorick@google.com
Labels: Needs-Feedback
jbudorick@, is there a document on how to reproduce errors found by this bot? I'm uncertain which commands it's running, and if there's a way to just run the commands that the bot runs, on the specific websites where it failed.

While the bot consistently hits these DCHECKs, I'm unsure which commands to run in order to trigger them, and can't seem to trigger them myself when messing around with the websites it hits DCHECKs on.
Cc: -jbudorick@google.com jbudorick@chromium.org
Sorry I used the wrong email. jbudorick@?

Sign in to add a comment