Pavel Emelyanov

OK, thanks for letting us know. I will change the .so name to something less "generic" :)

Check for retcode is not required there actually. The read() will not read in pre-initialized variable. If write() fails, there's nothing to be done about it :)

You're welcome ;)

https://app.circleci.com/pipelines/github/tchaikov/seastar/13/workflows/2924b836-4958-46e7-bd26-19ee5916fa2c/jobs/138 ``` ==523==WARNING: ASan doesn't fully support makecontext/swapcontext functions and may produce false positives in some cases! .../tls_test.cc(90): error: in "test_x509_client_with_builder_system_trust_multiple": check result has failed .../tls_test.cc(91): error: in "test_x509_client_with_builder_system_trust_multiple": check...

https://app.circleci.com/pipelines/github/tchaikov/seastar/15/workflows/38ca33cb-41e9-40d8-b4f5-eeca56260f04/jobs/162 ``` ==524==WARNING: ASan doesn't fully support makecontext/swapcontext functions and may produce false positives in some cases! Reactor stalled for 32 ms on shard 0. Backtrace: 0x47394a 0xc4becd 0xc4b355 0xa976f7...

> I also don't understand why my PR closed this issue I believe it didn't. The "closed" mark is next to the PR mentioning this issues, the issue itself has...

> I don't suppose anyone managed to get this issue locally? Nope. Pure cirecle-ci thing so far but happening frighteningly regular

> 1.) Maybe flush poller should be at least possible optionally to disable for a stream? It can be done with `output_stream_options.batch_flushes = false`, but socket code sets it to...

> But we can't really push verification into separate, pre-empted, threads either - they are way to many Why can't we make it in the thread-pool's thread? Even if there...

The 1st patch doesn't belong to this PR, please, submit it separately