GitLab

At least for some logging. This eliminates having more real parameters
than placeholders for the used messages. Other occurrences of the
function are preserved, as the corresponding messages have full set of
placeholders.

We want to disable the logger if there was an exception while putting
the parameters in. While we can't detect all possible ones, we can
detect some and reduce some false positives of the strict log checker
and not output mangled messages.

When we enable the strict log checking and the excess placeholders check
triggers, we need to print the message to easily identify the message ID
with the problem.

The DATASRC_DATABASE_FINDNSEC3_COVER logging message was missing a
placeholder. Added.

to reduce the risk of using already-installed older BIND10 libraries.
basic approach was already agreed.  this particular fix should be trivial.
so directly committing.

build on openbsd fails without this.
should be trivial, so commiting this at my discretion.

this fixes linker error on some system that has installed an older
version of BIND 10 in a well known place for the system, like /usr/local/lib.
okayed on jabber.

on macmini the compiler seems to apply the copy constructor and complain
about it.  that rather seems to be a compiler problem, but this small change
works around it.

committing it at my discretion.

- use recipient instead of forwarded as it seemed more intuitive
- descirbe how the request is responded
- use 'component' instead of 'program' where it's more appropriate
- use 'process' instead of 'program' where it's more appropriate

the formatEndpoint() function locally defined in auth_srv was essentially
moved there.  also added the test of the operator.

this will make the caller side much simpler.  in particular, the caller
now doesn't have to handle exceptions.  connect() and close() can now
be private.

To handle various types of requests, the constructor is extended to take
a message type, which will be used in the log message on failure.
log message ID and description were made generic.

so that it matches latest implementation status (and the statement is
generic and won't have to be updated this way again)

so that it doesn't necessarily mean we'll get the path from a configuration.

the latter isn't very portable.

added comments about the underlying assumption and what we should do
if that isn't held.

mainly covering some unusual setups.  no need to update the tested code.

this is the real fix to the distcheck failure on some systems.
in the end LIBRARY_PATH_PLACEHOLDER didn't only unsolve the problem, but
also unnecessary at all.  So the previous change on the makefile was
canceled completely.

for now, the effect is the same, and we may want to move the return
back to the common part if we want to handle multiple exceptions
in a unified matter.  But at the moment we don't have the need for it,
and the revised position seemed to be less confusing.

(the intent of the changes should be obvious from the diff.)

this will hopefully fix distcheck failure on some systems (will check).

otherwise link would fail for (e.g.) MacOS X with -no-undefined.
discussed on jabber, behavior confirmed by myself.

Asynchronous communictaion calls have been added

This is the asynchronous read on ModuleCCSession.

not really useful at this point anymore

It isn't actually used by the test, but by not setting any db, the default db would've been created (and not cleaned up, making distcheck fail for leaving files around)

As this is the better way reportedly.