[Tarantool-patches] [PATCH] coio: handle spurious wakeup correctly

Vladislav Shpilevoy v.shpilevoy at tarantool.org
Thu Nov 18 02:11:09 MSK 2021

>> On 16.11.2021 11:17, Serge Petrenko via Tarantool-patches wrote:
>>> coio_accept, coio_read, coio_write, coio_writev used to handle spurious
>>> wakeups correctly: if the timeout hasn't passed yet, they would simply
>>> retry reading (or writing) and fall asleep once again if no data is
>>> ready.
>>> This behaviour changed in the following patches:
>>> 577a640a7fdec986d19101ed04d2afa80e951c78 ("coio: pass fd to
>>> coio_accept") and 4f84859dcdd6126b0bdcda810b7f5f58386bdac6 ("Introduce
>>> iostream wrapper for socket I/O").
>>> Now the functions timeout on the very first spurious wakeup.
>>> Fix this, add the appropriate unit tests and a test_iostream
>>> implementation for the ease of testing.
>> Don't we have the same problem with coio_connect_addr() (used in
>> coio_connect_timeout())?
> Not really. Neither coio_connect_addr() nor coio_connect_timeout() retry
> the connection. So even the previous version would throw an error after
> a spurious wakeup. Just the error would be different.
> Before the change it would throw SocketError, not TimedOut, but I don't think
> this matters much. Does it?
> By the change I mean (2db0741b) "coio: return fd from coio_connect".

Ok, then never mind. I will wait for fixes of Vova's comments regarding the
tests, but the bugfix itself LGTM.

More information about the Tarantool-patches mailing list