On 26/02/2025 12:04, Mark Thomas wrote:
On 26/02/2025 08:16, Mark Thomas wrote:
On 13/02/2025 10:04, Rémy Maucherat wrote:
On Thu, Feb 13, 2025 at 9:41 AM Cenk Pekyaman
<cenkpekya...@gmail.com> wrote:
We run tomcat on java17 with the embedded tomcat setup.
We have http and https connectors and we have http2 upgradeProtocol for
both.
We recently upgraded from 9.0.88 to 10.1.24 to work on javax to jakarta
migration, and after a while, upgraded to 10.1.33.
After the upgrade, we started to see random and rare 500 errors for
some of
the http2 GET requests over https.
When reproducing the error on our development machines, we could see
the
following trace on the server:
```
java.io.IOException: null
Tomcat now sets an IO exception to trigger ReadListener.onError with
an appropriate error in that case. So things seem normal so far.
Following up on this.
Tomcat is behaving as if the client has reset the stream before the
client sent all of the data. If that is what the client is doing then
this behaviour is expected. However, that then raises the question why
is the reset being sent. If the client isn't resetting the stream then
there is definitely a Tomcat bug here. We need to look at the test
case you've provided. I'm planning on starting that today.
Thanks again for the test case. It makes debugging issues so much easier
when the report includes a reproducible test case.
I am able to reproduce this issue with the latest 12.0.x code.
It does look like there is a Tomcat bug here. I am currently working on
tracking down the root cause.
Found it and fixed the issue. The fix will be in the March releases.
Mark
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
