[
https://issues.apache.org/jira/browse/BEAM-9382?focusedWorklogId=399325&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-399325
]
ASF GitHub Bot logged work on BEAM-9382:
----------------------------------------
Author: ASF GitHub Bot
Created on: 06/Mar/20 19:45
Start Date: 06/Mar/20 19:45
Worklog Time Spent: 10m
Work Description: robertwb commented on issue #10988: [BEAM-9382] Clean
up of TestStreamTranscriptTests.
URL: https://github.com/apache/beam/pull/10988#issuecomment-595930667
The Python direct runner does support PaneInfo, but it doesn't (yet) support
allowed lateness, so the "final" bit is wrong. Yes, we should fix this, though
hopefully on the FnApiRunner once it supports streaming rather than invest too
much into the old direct runner (unless it's cheap to do).
I would rather disable the "final" bit check for these tests (and add a new,
disabled test that checks this bit) than disable these tests entirely for the
direct runner. So as is, this change LGTM.
We could also consider expanding the API to indicate certain features are
broken on certain runners, rather than entire tests, which would allow us to
fully specify the expected results and still produce partial validation on
incomplete/in-progress runners.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 399325)
Time Spent: 1h 10m (was: 1h)
> TestStreamTranscriptTest relies on non-deterministic behavior
> -------------------------------------------------------------
>
> Key: BEAM-9382
> URL: https://issues.apache.org/jira/browse/BEAM-9382
> Project: Beam
> Issue Type: Bug
> Components: testing
> Reporter: Andrew Crites
> Assignee: Andrew Crites
> Priority: Minor
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> The test discarding_early_fixed uses an early trigger Count(2) and then
> inserts 3 elements, assuming all 3 will get emitted in the early pane.
> However, runners do not have to follow this behavior. Instead, they could
> emit the first two elements seen and then buffer the third until something
> else comes in. We should change this test to only insert 2 elements so that
> all runners will behave the same.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
