Am 24.09.2018 um 14:27 schrieb Mark Thomas:
On 24/09/2018 05:36, Rainer Jung wrote:
What should we do in the light of BZ 62751 (r1841817)? It is a
regression in 1.2.44 when the timeout based on poll() was added.
The bug means, that a short connect (=connection setup) timeout
someone might have configured actually results in a long timeout
(using seconds instead of milliseconds as expected and documented).
By default no timeout is set, so the bug hits only, when a
socket_connect_timeout was explicitly set (and the platform is one
that has poll() and is not Windows; that should be true for most
non-Windows platforms). The example config that is part of the source
distribution and was suggested frequently (and least by me) does set
the timeout though:
# - socket_connect_timeout: milliseconds, default=0
# Since: 1.2.27
worker.template.socket_connect_timeout=5000
This should behave like 5 seconds, but now behaves like about 1.5
hours timeout.
Since it is a 1.2.44 regression, should we drop 1.2.45 and do a 1.2.46
with only that fix on top?
I think that makes sense but I don't have access to the Windows VM I
normally use to do this as I am at ApacheCon this week. I might be able
to work around that though.
I'll cancel the vote shortly and see about rolling 1.2.46.
Mark
Thank you and have an interesting conference!
Regards,
Rainer
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
