Varnish Enterprise 6.0.18r4 is a bug fix release with no new features. It contains important stability and correctness fixes, and all users are encouraged to upgrade.
Fixed a bug where vcl.discard could make the manager process exit if
the child process failed to answer the CLI in time, leaving the cache
stopped until manual intervention. The manager now reports the failure
and restarts the child.
Fixed a crash in VSM clients (varnishstat, vmod_stat and other log
and stat consumers) where a shared memory segment cluster could be
freed while still in use. (VS issue #3252)
Fixed a cache lookup edge case where a soft-purged object still within its grace period could be fetched synchronously as a miss instead of being served stale with a background refresh. (VS issue #3218)
Fixed a case where the HTTP client IP could be appended twice to the
X-Forwarded-For header when using the vmod_headerplus function
write_req0(), as used in the root VCL of Varnish Controller
deployments. (VS issue #3309)
varnishd no longer panics at startup when launched with the standard
file descriptors closed.
A number of request body handling fixes from Varnish Cache, improving
Content-Length consistency after unset bereq.body and returning a
400 response status on request body read failures.
See the changelog for the complete list of changes.
