These were only manual due to different build requirements.  This
adds a new "xtest" category, which is included in the automatic tests
but meets the requirements for the darwin_c tests.

TESTED:
The darwin_c test are now included in the "test" target.

It's more consistent with normal naming conventions to use a
leading underscore for the name of this flag.  Thus, the
flag is now _MACPORTS_LEGACY_MIN_EARLY_SDK_ALLOWED.

Since this feature is only recently added and rarely needed, it's
not too likely that this change will break anything.

TESTED:
Updated name has desired effect.

Apple failed to update AvailabilityMacros.h for the 15.0 SDK, so it
appears to be a 14.x SDK.  This adds a workaround for that.

TESTED:
The checksdkversion test now reports correct results for all available
SDKs (10.4-15.0).

This adds a wrapper to make the scandir() 'compar' argument and the
provided alphasort() function match the types used in 10.8+, thereby
avoiding warnings, and in some cases errors, due to the pointer type
mismatch.

To avoid breaking code expecting the old behavior, this hack is only
applied when _MACPORTS_LEGACY_COMPATIBLE_SCANDIR is defined nonzero.

Closes: https://trac.macports.org/ticket/70702

TESTED:
Tests pass on all OS versions with this fix.

This tests for the pointer-type mismatch issue with the scandir()
'compar' arg.  There are two versions, one verifying the expected old
behavior, and another verifying the fix.

See: https://trac.macports.org/ticket/70702

Also adds alphasort() and scandir() to the darwin_c tests.

TESTED:
Fails or gets warning on <10.8 without fix.  Passes with fix.

TESTED:
Fails or gets warning on <10.8 without fix.  Passes with fix.

This is primarily to fix a bug in dprintf() which erroneously closed
the fd after its use, but is also mildly more efficient.

Based on some work by raf <[email protected]>

TESTED:
Passes the updated test, which failed before this fix.

This performs a much more thorough test of dprintf(), including
verifying that it does not close the underlying fd.

TESTED:
Fails when the old close bug is present, passes otherwise.

Also rewrites dprintf() to be a wrapper around vdprintf(), rather than
duplicating the "guts".

Based on some work by raf <[email protected]>

TESTED:
Passes updated test.

Also adds vdprintf() to darwin_c tests.

TESTED:
Fails without the new vdprintf() addition; passes with it.

This uses a kludgy and nonportable method to avoid the ill effects of
fclose() on locks (see the comment).  It should be reliable for all
relevant OSX versions, but it can be reverted if it causes trouble.

TESTED:
Passes tests, which would fail if the close() disable didn't work.