tree 5f3889eae2566bbeb3741e2391e60f4b50bf8a50
parent 445672132889833b4a26d7df869b429962c2af45
author Kannan J <kannanjgithub@google.com> 1781869308 +0530
committer GitHub <noreply@github.com> 1781869308 +0530
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsFcBAABCAAQBQJqNSr8CRC1aQ7uu5UhlAAALJoQAD4JkNRSrLqw5EP8X87d8ZVA
 Hig1BToAhebmp+8GIkIdkM15mhGfQHPtJlM5phXa6AXj7M/RaXRU+0Ft8haa+1ra
 ylddIs9RHWTGiZLiB9EDvzAK+6cVxRWg5Juzfb7yzlGi4srAttlU5Wrbe4KZ3fGi
 lsUwqyoLK71/MEq83sVtRFL5voAeB9J5MMNn7y306JrybHx8K6FP3wbxNgqSfhKf
 nD424GuyV5LR7fsIse0i7Jywnw93Oh5q2HpqBmjJ2xANTq+KHnfip5ulZTlLRRXO
 iZY9Khq7SE6wfCoZ2+k2xoz+PCCxsRLj4m/zUmKteGfFxK/89yjuipxJEycg1qOE
 OlP/WVkrvQXD1qaeMCNTF4BP+wWI0lyitEtTocKC89akiYLJ+CUw+/btcgMl7oRu
 D0ex21bXSIy7aPJlpP2x1mcxwC08qEHi2hatkOkgmdEcqiwHW8hjlTQSJdnKau4b
 kjR3zMKr8pejrLlfZDewhXwxYMCA0GVjmemH6qhG99HzIpT8PwM6B57YqZmcdCBr
 FkwjSsrez+nUYJ2N8PbSKuue4tp/0sFWbYet4+DBr+ZTS7BW4ztJBctvit5bywbu
 nFUeRr7Ie36DruT5ogQYYjLlD9R6Fzyf8PFSLWGS/yplW/Rij1G3D7EGmE1STyCj
 7Nf6S7ZCktSW0vevy16W
 =BpTn
 -----END PGP SIGNATURE-----
 

Revert "servlet: fix write when not ready in AsyncServletOutputStreamWriter (#12867)

This reverts commit 243b79ee5b2419ac8a71e3870499ec84233972d2. The
approach taken of ensuring subsequent writes go through the container
thread via onWritePossible() hasn't eliminated the need for catching the
IllegalStateException (which you are catching and buffering the write if
it happens). We have observed that the UndertowInteropTest now hangs
instead of the earlier flaky failures due to uncaught exception. There
has been another PR https://github.com/grpc/grpc-java/pull/12732 for the
same issue that explicitly checks isReady.getAsBoolean() in addition to
the cached state, that's probably most reliable, because you don't have
to deal with callback scheduling and the state changing in the meantime.
We would like to proceed with that PR and revert this one.

@mgustimz