Hi Markus,

Thank you, your patch does indeed fix the build problem for this package, however, the test failure is not insignificant.

I did further tests with puppetserver, which is a downstream dependency of trapperkeeper-webserver-jetty9-clojure and unfortunately, the web requests (access) logging remains broken. There are no warnings or error messages anywhere: as you can imagine, the logging events are simply lost in the ether.

I think these logs are a critical feature for debugging and security and it would be very unfortunate for bookworm to ship with logging crippled in this way for puppetserver.

Although I did not test this specifically, I strongly suspect puppetdb is also affected in a similar way because they share many of the same components.

Do you think it might be possible to backport the accessEvent variable fix from a later version of Logback?

-- Jérôme


Le 2023-05-19 à 09 h 53, Markus Koschany a écrit :
Control: tags -1 patch

Hi,

I am attaching a patch for this issue. Somehow your package fell through the
cracks because it has no dependency on logback. Logback is pulled in by Jetty
though. While Logback and Jetty use the Jakarta servlet API now, your package
still depends on libservlet-api-java. The fix for that was trivial. However the
test failure is caused by my tomcat10-migration.patch

https://salsa.debian.org/java-team/logback/-/blob/master/debian/patches/tomcat10-migration.patch#L91

Back then I thought it was acceptable to work around a build failure by setting
the accessEvent variable to null since it seemed no package existed in Debian
which would be affected by it. The Logback developers fixed that problem in
newer versions, but a new upstream release of logback would have been too
intrusive. Long story short: please double-check my patch and if we can ignore
the logging problem

Regards,

Markus



Reply via email to