Before C11, repeating a typedef is forbidden, if taken strictly, and
some compiler versions even reject it.  On the other hand lru.h was
missing it, which wasn't noticed due to glibc's <sys/types.h> defining it.
Fixes https://gitlab.labs.nic.cz/knot/resolver/issues/114.

Vladimír Čunát authored 8 years ago and Ondřej Surý committed 8 years ago

Motivation: maintaining these by hand in kres.lua is tricky, as just
inserting or reordering the structures can do anything and isn't
detected in general.  For example, `rr_array_t::at` has always
"missed a star".

A couple libknot structure fields have apparently changed name since
the old definitions were written; in most cases I renamed them in *.lua.

The file daemon/lua/kres-gen.lua will be regenerated only if explicitly
deleted.  That requires building with debug symbols and gdb on $PATH.

... fixing a warning.  The addresses are the same,
but it wasn't completely correct.

Now that we renamed "debug" to "verbose", the old name didn't seem very
suitable anymore; this new identifier haven't got to master yet.

Split NDEBUG to NDEBUG (asserts) and NOVERBOSELOG (verbose logging); rename all related symbols to use VERBOSE instead of DEBUG

In particular, don't require to call an externally defined function to
find if in --verbose mode or not.  Now it's just an extern bool.
I'm confident the performance impact of not using -DNLOGDEBUG should be
negligible now.

This comes with a small incompatible API+ABI change, but that shouldn't
matter as we've had a couple of those already since the last release.

Fixes #111.

Compatibility: it needs a new libkres.so function.  If a wrong version
combination is attempted, the module just cleanly fails to load, though
it writes a slightly confusing message "no such file or directory".

- Drop all make dependencies, so it's only generated when the file
  goes missing (i.e. was explicitly deleted).
- Add checks into the generator to avoid producing empty strings on
  failures.  It compiled correctly and wasn't easy to debug.

We only want this when deleted explicitly.

- The API and ABI for modules changes slightly (details below).
  KR_MODULE_API is bumped to avoid loading incompatible code.
  We have bumped libkres ABIVER since the last release 1.1.1,
  so leaving that one intact.

- Make KR_STATE_YIELD not reuse 0 value anymore.
  It's easy to e.g. return kr_ok() by mistake.
- struct kr_layer_t:
  * ::mm was unused, uninitialized, etc.
  * Make ::state an int, as it was everywhere else.
  * void *data was ugly and always containing struct kr_request *
- struct kr_layer_api:
  * Drop the void* parameter from ::begin, as it was only used
    for the request which is available as ctx->req anyway
    (formerly ctx->data).
  * Drop ::fail.  It wasn't even called.  Modules can watch for
    KR_STATE_FAIL in ::finish.
- Document the apparent meaning of the layer interface, deduced mainly
  from the way it's used in the code.  Caveats:
  * enum knot_layer_state handling seems to assume that it holds exactly
    one of the possibilities at a time.  The cookie module does NOT
    follow that (intentionally), apparently depending on the exact
    implementation of the handling at that moment.  It feels fragile.
  * I was unable to deduce a plausible description of when ::reset is
    called.  It's practically unused in modules, too.

It causes lots of line changes, but it would be confusing to keep the
current state over long term.

Fixes #100.
The file with the addresses is in the repo, so people don't have to
regenerate it, but the build system is able to do it.

I don't know why it passed on Linux without any warning.

... and that doesn't necessarily mean that malloc() failed.
We do *not* want to evict a heavy-hitter by an unfrequent element.

Note: even the implementation currently in master *did* return NULL,
so some parts of the code were just wrongly returning ENOMEM.

The implementation is now similar to set-associative caches
that x86 CPU use.  Also the API is changed a bit, leading to
slight simplification of our use patterns.

Fixes https://gitlab.labs.nic.cz/knot/resolver/issues/93.
API of a KR_EXPORT function is changed, so ABIVER is bumped.

if the score is known (and not glued), the retry
rate is <avg(rtt), 250ms>

if the NS score is unknown or worse than 250ms,
it is always retried at this rate

all other servers in the list use default rate

maintainer may wish to choose neutral or even
more favoured server selection of IPv6 servers.

-DFAVOUR_IPV6=0 makes neutral server selection
-DFAVOUR_IPV6=40 gives IPv6 40ms better chance

RTT tracking for all targets is also supported,
but no loadbalancing is done based on that yet

when doing fast retransmit, each address may be
contacted 1-N times, but previously only cumulative
RTT was tracked for the NS that sent the answer.

now the approximate query start time is subtracted
from cumulative RTT to give an idea how long it
took since the query was actually sent, and all
the NSs that didn't respond within their retransmit
windows are penalised too