[pkg/stanza] Fileconsumer TestFlushPeriodEOF test failing on Windows

[crobert-1]

Component(s)

pkg/stanza

Describe the issue you're reporting

Failing CI/CD links:
https://github.com/open-telemetry/opentelemetry-collector-contrib/actions/runs/8849403200/job/24301322178
https://github.com/open-telemetry/opentelemetry-collector-contrib/actions/runs/8850888073/job/24306184699?pr=32529

Failure output:

=== Failed
=== FAIL: fileconsumer/internal/reader TestFlushPeriodEOF (3.01s)
    sink.go:114: 
        	Error Trace:	D:/a/opentelemetry-collector-contrib/opentelemetry-collector-contrib/pkg/stanza/fileconsumer/internal/emittest/sink.go:114
        	            				D:/a/opentelemetry-collector-contrib/opentelemetry-collector-contrib/pkg/stanza/fileconsumer/internal/reader/reader_test.go:209
        	Error:      	timeout: expected: 2, actual: 1
        	Test:       	TestFlushPeriodEOF

=== FAIL: fileconsumer/internal/reader TestFlushPeriodEOF (re-run 1) (3.02s)
    sink.go:114: 
        	Error Trace:	D:/a/opentelemetry-collector-contrib/opentelemetry-collector-contrib/pkg/stanza/fileconsumer/internal/emittest/sink.go:114
        	            				D:/a/opentelemetry-collector-contrib/opentelemetry-collector-contrib/pkg/stanza/fileconsumer/internal/reader/reader_test.go:209
        	Error:      	timeout: expected: 2, actual: 1
        	Test:       	TestFlushPeriodEOF

DONE 2 runs, 3289 tests, 13 skipped, 2 failures in 112.066s

This has been failing pretty consistently (but not quite everytime) for the last few days, I haven't been able to determine why it recently started failing more frequently.

[github-actions]

Pinging code owners:

• pkg/stanza: @djaglowski

%s See Adding Labels via Comments if you do not have permissions to add labels yourself.

[djaglowski]

Could this be related to #32100? @OverOrion @ChrsMark

[OverOrion]

I think it is highly likely to be related. I'll try to have a look at it this week, not sure if I can before SIG though.

EDIT: maybe it has something to do with using \r\n instead of \n?

[ChrsMark]

maybe it has something to do with using \r\n instead of \n?

That's likely :) (windows 🙈).
Maybe using the the fmt library to add the new line (ie fmt.Sprintln) instead of adding it explicitly could help here?

[ChrsMark]

So for some reason on Windows the second token is not collected if we don't append a new line (\n) after it.
That's why we get the error expected: 2, actual: 1, because we were reading only 1 token instead of 2.

It seems to be fixed at #32734.
I'm trying to think though if this error indicates that there is an actual issue in the implementation with how we check for the EOFs.
@djaglowski @OverOrion would that be possible?
Do we have other tests already that would ensure the behavior on Windows?

[crobert-1]

+1 freq: https://github.com/open-telemetry/opentelemetry-collector-contrib/actions/runs/8940807269/job/24559739110

[crobert-1]

+1 freq: https://github.com/open-telemetry/opentelemetry-collector-contrib/actions/runs/8939477614/job/24555590652

[crobert-1]

+1 freq: https://github.com/open-telemetry/opentelemetry-collector-contrib/actions/runs/8993044783/job/24704110759?pr=32932