Deal with AIX defining in6addr_any but not in6addr_loopback
in libc, even though both are declared in <netinet/in.h>. This caused
named to fail to link on AIX after change 1083, which introduced
the first actual use of in6addr_loopback.
1068. [bug] errno could be overwritten by catgets(). [RT #1921]
1066. [bug] Provide a thread safe wrapper for strerror().
[RT #1689]
(Also some changes to configure.in, config.h.in, and acconfig.h that were
necessary to allow config.h.in to be correctly regenerated with the
HAVE_STRERROR definition needed by 1066.)
1021. [bug] On Win32, log message timestamps were one month
later than they should have been, and the server
would exhibit unspecified behavior in December.
doio_recv was not catching a ECONNRESET error. Fixed this and added code to catch other possible errors. Miscellaneous errno handling cleanup [RT #1745]
[RT #1709]
Change isc_entropy_usebestsource() to have saner semantics:
- If an invalid file is specified, an error will be returned instead of the
keyboard being used.
- If no file is specified but a random device is present, the keyboard will
be used if there is an error opening the random device.
- ISC_ENTROPY_KEYBOARDYES indicates that the keyboard should be the
only device used. Otherwise, passing '-r keyboard' is meaningless
on a machine with a random device, since the keyboard will not be used.
Change the callers in the dnssec tools and rndc-confgen to check for the
special file "keyboard" and call isc_entropy_usebestsource() with the right set
of parameters.
972. [bug] The file modification time code in zone.c was using the
wrong epoch. [RT #1667]
(This pullup includes not only the changes to zone.c, but all the
recent epoch-related fixes)
RT #1675 resolv.conf was specified to be in an invalid directory.
Location was changed to be in the same etc directory as all of the other
configuration files
'magic' was not being declared consistantly.
some #include <isc/magic.h> were missing from other include files.
NS_SERVER_VALID was not using ISC_MAGIC_VALID.
