https://bz.apache.org/bugzilla/show_bug.cgi?id=63765
Remy Maucherat changed:
What|Removed |Added
Resolution|--- |FIXED
Status|NEW
https://bz.apache.org/bugzilla/show_bug.cgi?id=63765
--- Comment #7 from Remy Maucherat ---
(In reply to Rainer Jung from comment #6)
> Can't say whether it is safe though.
It likely sounds safe to unwrap first (it would do an underflow and cause a
read) reading some previous comment:
https://gi
https://bz.apache.org/bugzilla/show_bug.cgi?id=63765
--- Comment #6 from Rainer Jung ---
(In reply to Remy Maucherat from comment #1)
> This looked like the usual not-unwrapping-enough "IO" issue, so changing the
> initial value of unwrapBeforeRead to true in SecureNio2Channel.reset avoids
> it.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63765
--- Comment #5 from Rainer Jung ---
And here Nio2 using TLS 1.2:
...
24-Sep-2019 15:16:55.232 FINE [https-openssl-nio2-8443-exec-4]
org.apache.tomcat.util.net.SocketWrapperBase.populateReadBuffer Socket:
[org.apache.tomcat.util.net.Nio2Endpoin
https://bz.apache.org/bugzilla/show_bug.cgi?id=63765
--- Comment #4 from Rainer Jung ---
For comparison the Nio (not NIO2) log snippet where the request works:
24-Sep-2019 13:54:40.068 FINE [https-openssl-nio-8443-Acceptor]
org.apache.tomcat.util.threads.LimitLatch.countUpOrAwait Counting
up[htt
https://bz.apache.org/bugzilla/show_bug.cgi?id=63765
--- Comment #3 from Rainer Jung ---
I can also reproduce on SLES 12 and RHEL 7 but not on Solaris.
For me it does not happen with a real remote connection, but always with local
ones and with connections from another virtual guest on the same
https://bz.apache.org/bugzilla/show_bug.cgi?id=63765
--- Comment #2 from Coty Sutherland ---
> I cannot reproduce this that reliably personally, it may be only visible on
> localhost.
Weird. I can reproduce this between two machines as well, so it doesn't seem to
be a lo only issue.
--
You ar
https://bz.apache.org/bugzilla/show_bug.cgi?id=63765
--- Comment #1 from Remy Maucherat ---
This looked like the usual not-unwrapping-enough "IO" issue, so changing the
initial value of unwrapBeforeRead to true in SecureNio2Channel.reset avoids it.
Not sure if it is 100% safe (although it looks r
