Walt Karas comments

Results 115 comments of


                                            Walt Karas

In event processing, assert that mutex is unlocked before freeing mutex.

[approve ci autest]

Plugins ought live in lib rather than libexec

I believe lib and libexec can be mapped to the same directory using runroot.yaml: https://docs.trafficserver.apache.org/developer-guide/layout/runroot.en.html#the-yaml-file

Use std::unique_ptr to replace "scoped" classes like ats_scope_fd and the MutexLock classes

@Mouray-Hutchinson are you still interested in working on this? What's your background in software development?

Stop write_fail actions from going to origin if parentage with go_direct=false

If possible, attach the ATS configs and HTTP message captures with last hop and next hop for when this occurs, or an Au test.

Do some core intercept plugins leak resources on early session termination?

Maybe it's only a problem if the plugin intercepts requests with content, (POSTs, PUSHes)? I would guess TS waits to receive the complete request header before calling any hooked plugin...

Do some core intercept plugins leak resources on early session termination?

ts_lua_http_intercept is a server intercept, so maybe the problem there would be if server sends no or incomplete data.

In ja3_fingerprint plugin, protect against premature VConn closure.

This is going into Yahoo Prod first. Will undraft this after it soaks in prod for a while.

In ja3_fingerprint plugin, protect against premature VConn closure.

This has soaked in Yahoo Prod for a couple of weeks, no issues. It potentially eliminates some crashes, but I'm not 100% it does.

In ja3_fingerprint plugin, protect against premature VConn closure.

This is a stack dump for a type of crash that we were seeing in Yahoo prod, that I hope this change fixes: ``` [ 00 ] libc-2.17.so raise (...

Fix an error on SSL config reload (plus some cleanup).

I'll undraft this after letting it soak in Yahoo Prod for a while.