fails.

require X+1 bytes where X is the largest of width, precision, and the
strlen() of the passed pointer.  If this is larger than MAX_ARG_LEN,
malloc() a buffer of size X+1.

sdl.Quit() to restore the console mode.

All further wrappers should be able to call xp_vasprintf() and Just Work
now imho.

wrappers can test this and stop reading va_args.

consitently on all platforms, so can't use it to convert a char * to a double.
(using strtod()).

Some of the conversions are dubious... ie: should a char * just be case to
an int, or atoi()ed to one?  Using the second method currently.

mess of switch()es... basically, each type will have to have each type.

as I'm unsure as to how widely supported they are.

converted to the "correct" type before replacement. (Not yet implemented)

string to prevent %s replacements from having format specifiers which get
replaced.

a number of non-x86 archetectures, most notably I64 and Linux/PPC.

Essentially, the problem is that there is NO way to portably contruct a
va_list at run-time.  Currently, this is done in (at least) execmisc.cpp
(and probobly in js_global.c or wherever format() and friends are)

This provides an alternate method of handling such issues which could
(potentially) solve the problem as well as add float/double support.
Comments are encouraged.  This is *NOT* part of the xpdev build and will
not be until it is tested and the marked ToDo items are dealt with.
Further, it most likely won't be *USED* in Synchronet until v4, or until
it recieves heavy testing on a very wide range of platforms and format
specifiers.

xp_*printf() functions for all the standard *printf() functions should be
written, including asprintf() and vasprintf() from the GNU C library.
This will require a new type argument (XP_PRINTF_TYPE_AUTO perhaps) which
will assume that the passed argument is the correct type.

and URANDOM_DEV used for seeding.
Fall-through seeding if there are problems opening URANDOM_DEV.
Use random() and srandom() where available as it's considered quite a bit
better than rand().
Define (but don't use) HAS_SRANDOMDEV_FUNC to indicate that srandomdev()
is avaiable.  The reason it's not used is that xp_randomize() returns the
seed value presumably to make it possible to replay a random sequence and
srandomdev() creates states which cannot be replicated on demand.

writes from JS stuff.

A half-decent start on implementing write_template() in C rather than JS.
Unfortunately, this started to feel too much like my day job, so it's
getting tossed on the pile for (at least) tonight... #ifdef'ed out.

-O3 causes problems on some Ubuntu.

Thanks RuneMaster!

from the ringbuffer but before it is actually sent.

close_request().  Drain the outbuf.

close the socket.

dynamically allocated (duh). Thanks DJCanDeeFlip for the bug report.

(rather than _DEBUG) - improving performance and memory usage in the 99% of
debug builds that don't need the JSDOC stuff.

And just have numbered notes.

Sure, we won't get as many usefull backtraces, but let's be reasonable.

sessions.

error.