The epoch indices are from [1..period].

 * changed stype to type
 * fix enquening queries from predict log
 * keep heuristic in function generate to predict upcoming epoch

... data from cache as keys for validation

... data from cache as keys for validation

Noticed after removal of its lua binding in commit 3be76db7.

Clearing the query flags lead to kresd iterating over authoritative
servers.  Related: https://gitlab.labs.nic.cz/knot/resolver/issues/217

Note that validation of synthesised AAAA behind a CNAME doesn't work
when policy.FORWARD + dns64.  Example: www.regiojet.cz.
policy.STUB + dns64 seems not to work at all.

If CNAME chain occurs CNAME is placed before AAAA.

Maybe some coverity issues will get fixed, even though
they had seemed not to be affecting our use cases in kresd.

Tests+Deckard look OK and I've been using my system and kresd with
system-wide 0.9.21 for some time already.

Closes #125 and #208.

- sort the list instead of just picking the best one
- prefer unknown RTTs to probe them
- verbose output of the choice

Fixes https://gitlab.labs.nic.cz/knot/resolver/issues/125
Fixes https://gitlab.labs.nic.cz/knot/resolver/issues/208

They're just useless when we have kr_inaddr*.

There would be no real resource leak, as the memory is on a mempool,
so the lifetime is limited by kr_request finishing.  Also, it's only
on an unlikely error branch, but still...

Some people might get surprised, as it's the default location for quite
some data nowadays ($XDG_DATA_HOME).

I just found it by accident when studying RFCs again.
Hmm, I'd like an operator for logical implication; I'm spoiled by Maths.

daemon: check existence of config file

See merge request !317

Without it we were failing to download fstrm tarball,
and the shown errors were very confusing.
Tested building an image locally.

... even if rundir isn't specified.  No other changes in semantics.
Before this a typo in config path would pass silently.

Grigorii Demidov authored 7 years ago and Vladimír Čunát committed 7 years ago

unecessary queries in some circumstances; some minor bugfixes

Grigorii Demidov authored 7 years ago and Vladimír Čunát committed 7 years ago

it can contain records which have not been validated by validator

Grigorii Demidov authored 7 years ago and Vladimír Čunát committed 7 years ago

Don't check proof of nonexistance in NODATA or NAME ERROR answers
which contains a CNAME.

In particular, don't mention things that probably don't work (well):
Windows, amalgamation, very old OpenSUSE package.