Revert "[BEAM-8111] Add ValidatesRunner test to AvroSchemaTest"

[markflyhigh]

Reverts #9454 due to AvroSchemaTest.testAvroPipelineGroupBy consistently failed in Flink ValidatesRunner test suite (link).

This is the first failure instance. You can go to console output or Gradle scan for error details.

+R: @reuvenlax@kennknowles@lukecwik

[reuvenlax]

FYI this appears to be Flink ValidatesRunner, not Dataflow?

…

On Wed, Sep 11, 2019 at 11:30 AM Mark Liu ***@***.***> wrote: @markflyhigh <https://github.com/markflyhigh> requested your review on: #9550 <#9550> Revert "[BEAM-8111] Add ValidatesRunner test to AvroSchemaTest". — You are receiving this because your review was requested. Reply to this email directly, view it on GitHub <#9550?email_source=notifications&email_token=AFAYJVNDZVD773KSV7FB743QJE2OFA5CNFSM4IVXMSBKYY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGOTSINQNY#event-2626738231>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFAYJVPTEZ3RFDFQEQSVIILQJE2OFANCNFSM4IVXMSBA> .

[reuvenlax]

Also it appears that Flink relies on all the Coders being exactly the same, which is far more restrictive than other runners. Is there another way for Flink to do this - maybe comparing the TypeDescriptor instead of the Coder?

…

On Wed, Sep 11, 2019 at 11:54 AM Reuven Lax ***@***.***> wrote: FYI this appears to be Flink ValidatesRunner, not Dataflow? On Wed, Sep 11, 2019 at 11:30 AM Mark Liu ***@***.***> wrote: > @markflyhigh <https://github.com/markflyhigh> requested your review on: > #9550 <#9550> Revert "[BEAM-8111] Add > ValidatesRunner test to AvroSchemaTest". > > — > You are receiving this because your review was requested. > Reply to this email directly, view it on GitHub > <#9550?email_source=notifications&email_token=AFAYJVNDZVD773KSV7FB743QJE2OFA5CNFSM4IVXMSBKYY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGOTSINQNY#event-2626738231>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AFAYJVPTEZ3RFDFQEQSVIILQJE2OFANCNFSM4IVXMSBA> > . >

[reuvenlax]

Also I suspect that #9446 will fix this.

…

On Wed, Sep 11, 2019 at 12:05 PM Reuven Lax ***@***.***> wrote: Also it appears that Flink relies on all the Coders being exactly the same, which is far more restrictive than other runners. Is there another way for Flink to do this - maybe comparing the TypeDescriptor instead of the Coder? On Wed, Sep 11, 2019 at 11:54 AM Reuven Lax ***@***.***> wrote: > FYI this appears to be Flink ValidatesRunner, not Dataflow? > > On Wed, Sep 11, 2019 at 11:30 AM Mark Liu ***@***.***> > wrote: > >> @markflyhigh <https://github.com/markflyhigh> requested your review on: >> #9550 <#9550> Revert "[BEAM-8111] >> Add ValidatesRunner test to AvroSchemaTest". >> >> — >> You are receiving this because your review was requested. >> Reply to this email directly, view it on GitHub >> <#9550?email_source=notifications&email_token=AFAYJVNDZVD773KSV7FB743QJE2OFA5CNFSM4IVXMSBKYY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGOTSINQNY#event-2626738231>, >> or mute the thread >> <https://github.com/notifications/unsubscribe-auth/AFAYJVPTEZ3RFDFQEQSVIILQJE2OFANCNFSM4IVXMSBA> >> . >> >

[markflyhigh]

It should be Flink. Thank you Reuven!

Robin already had #9541 to skip this test and test is done. I think we can continue with #9541 and ignore this rollback.

[robertwb]

So should this be closed?