Do not set thread priority to realtime as it causes hangups on iOS single core devices. Proper fix pending.

it breaks re-attaching, needs a better fix (possible race condition here)
Fixes #539110

This reverts commit 88a3c6ea.

[QNX] socket accept gets interrupted by real-time signals.  Loop over ac...

[QNX] socket accept gets interrupted by real-time signals.  Loop over accept function if failure cause was a signal interrupt

	* method-to-ir.c (mono_method_to_ir): Check for null
	on delegate construction so closed-on-null is not allowed
	when using the delegate .ctor.

	If the delegate been constructed is open-on-target then
	the null check must not be applied as the IL pattern is:

	ldnull
	ldftn void Bla::Foo(int)
	newobj Bar (object, native int) // delegate void Bar(Bla,int)

Conflicts:
	mono/mini/method-to-ir.c

Don't believe its lies.

This reverts commit 2feb13aa.

Enable thumb support for BB10, improve NDK auto download to use internal...

Don't push threads which are not suspended on Win32. Fixes a race condition described on GC mailing list where a new thread is being created during suspension process. Should avoid another cause of GetThreadContext failues.

shared handles; don't install atexit(handle_cleanup) if shared handles are disabled (atexit() is called at process termination; not at dlclose() - which fails on platforms with a dynamically loaded mono lib)

bsearch behavior is undefined when 'key' param is a null-ptr (on android 'undefined' sometimes means calling abort)

ARM: Fix issue when passing many and big value types (structs) from native code to managed code as method arguments, would cause memory overwrite and crash later. The change more properly estimates space required to store arguments in DynCallArgs.regs[] array. (Case 502916)

Add GC_stop_world_external and GC_start_world_external that properly disable signals and acquire GC locks before calling GC_stop_world and GC_start_world respectively. Use these new functions external of GC in liveness code. May fix case 530367.

	The timeout was not being used in GetResponse.
	Fixes bug #603794.

	In some weird cases socket is null when we try to close it.

	Send a Set-Cookie header per cookie instead of a Set-Cookie2 with
	multiple cookies.

	Fixes bug #667550.

	This fixes a bug reported in stackoverflow and mono-list that affects
	Mono 2.8 and Mono 2.6 when using BeginReceive/BeginSend overloads that
	take an ILIst<ArraySegment> argument.

	When Flush() is called explicitly, do a Z_PARTIAL_FLUSH to force gzip
	into giving us whatever is available. Less efficient, but compatible
	with MS.

	HttpListenerResponse could be calling HttpConnection.Close() more than
	once.

	If the Set-Cookie header is empty, the cookie instance is not created.
	Fixes bug #670189.