* gandi/livedns: Print actual changes to be pushed
Currently, preview & push output prints all domain records values:
$ dnscontrol preview
----- DNS Provider: gandi...1 correction
#1: Setting dns records for example.com:
mail.example.com.
1.2.3.4
"ga"
"bu"
"zo"
With this change, it would only show what changes from current state:
$ dnscontrol preview
----- DNS Provider: gandi...1 correction
#1: Setting dns records for example.com:
CREATE TXT ga.example.com "ga" ttl=10800
MODIFY TXT bu.example.com "bu" ttl=10800
DELETE TXT meu.example.com "meu" ttl=10800
* Add import
* condense AD 'unsupported record type' warnings
* active directory provider can create/modify/delete NS records!
* ad passes more of the tests now
* skip root NS records for safety
* clarifying comment
- Support DelegationSet for Route53 (create-domains only)
- Retry Route53 operations which fail for rate limits under large numbers of domains
- Support for name_server_set for GCloud (create-domains only)
- Docs for both
* Maint: run generate for missing documentation
Apparently current master is missing some generated documentation.
* Populate ovh zones cache as early as possible (#412)
We are caching the OVH zones in GetNameservers.
It turns out it isn’t a good idea, because GetNameServers will not be called
if the user selects no name servers for a given domain by using for example:
```
D(‘my domain’, DnsProvider(ovh, 0)) {
}
```
The subsequent GetDomainCorrections would automatically fail
with an unknown domain error, because the zones cache hasn’t been
filled in.
To solve the issue, the ovh provider now populates the zones cache during
initialisation.
This:
* adds a global -v flag for verbosity
* refactors the "printer" package to have a DefaultPrinter and package
functions that call it, similar to net/http's DefaultServeMux
* adds printer tests
* moves current users of Debugf to Printf
* moves most users of the "log" package to use "printer"
* demotes noticably noisy log messages to "Debugf", like "IGNORE"-
and "NO_PURGE"-related messages
Previously, unnecessary corrections were possible if both an R53_ALIAS
pointing to an A record and to an AAAA record existed for the same label,
and map iteration over existing and desired found them in different orders.
(This is a common configuration for IPv6-enabled records.)
This commit:
* mirrors key logic in the R53 provider
* centralizes logic around keys in the models package
* adds tests
