GitLab

This fixes memory leak reported by Valgrind. Dicussed with Jelte on jabber.

Strictly, this is done to silence clang-analyzer. In every case,
parseAnswer() does set rcode (which is passed by reference) unless it
throws.  It looks like clang-analyzer gets confused by the exception
paths.

In any case, it's good to initialize it to the unexpected case and not
depend on parseAnswer().

The assert() should silence clang-analyer.

These are not required according to:
http://docs.python.org/py3k/extending/extending.html?highlight=null#null-pointers

These NULL checks cause static analysis reports in other parts of the
code where the pointers are not NULL tested before use.

Reported by clang-analyzer.

Discussed with jreed on jabber.

In order to run this test interface detection is required. Since it is
right now implemented on Linux only it had to be disabled on every other
OS.

This is supposed to resolve ambiguous uintX_t types on Solaris 11.

I just happen to notice it, and since it's very trivial I'm making the
change directly.

This should fix the valgrind error report, verified on jabber

It doesn't happen currently. See ticket  http://bind10.isc.org/ticket/2240

We have had 34 random off-by-one failures on real and virtual machines.

Previously it allowed a two second range (it slept for 2 seconds).
Now allow a three second range.

Discussed on jabber.

be consistent

be consistent

discussed with vorner via jabber.

Check if socket has not been closed after timeout in read

discussed on jabber