[Nasm-devel] variadic single-line macros: want feedback

[anonymous c. <nas...@ya...>]

I have managed to add support for variadic single-line macros
to NASM, and I'd like to gather some feeback before I proceed.

1. Naming

While other languages use the term "variadic" to describe the
ability to lump more than N arguments into the Nth argument,
the traditional NASM term -- used to describe this feature in
the context of multi-line macros -- is "greedy". I'd like to
settle on one term, throughout all of NASM.

2. Syntax

While C uses "#define f(one,...) one,__VA_ARGS__", NASM does
use e.g. "%macro m 1-2+" (i.e. a trailing "+") for its greedy
multi-line macros. As of writing this, my implementation is
using "%define f(one,...) one,...", i.e. it does resemble C,
though it uses ... instead of __VA_ARGS__. However, I could
easily change this to "%define f(one,two+) one,two", which
would resemble NASM's multi-line macro syntax. I could also
implement "%define f(one,two...) one,two"; however, doing so
would be slightly more complicated.

3. Overloading

Due to NASM's ability to overload macro definitions, there is
no need for implementing equivalents to the two GNU GCC CPP
extensions that allow omitted variable arguments, as well as
the special ## treatment. In NASM one can just do

  %define eprintf(format)     printf(format)
  %define eprintf(format,...) printf(format,...)

instead (assuming my aforementioned syntax). For additional
details on this issue, please read the GCC CPP manual, i.e.
http://gcc.gnu.org/onlinedocs/cpp/Variadic-Macros.html.

That said, my current implementation merely adds a new boolean
flag to the definition of an smac, which indicates whether or
not that smac is variadic (and which of course causes it to be
handled accordingly in expand_smacro()). As a result, it does
behave as follows:

  %define smac(...)
  %define smac(one,two) ; no warning, now handles 2-arg case
  %ifdef smac(one,two,three)
   ; though smac(...) would handle this case, %ifdef does not
   ; consider a 3-arg smac to be a match for a 1-arg smac(...)
  %endif

I'm perfectly happy with the silent overloading; however, I'm
still debating the %ifdef issue (which also exists for %undef).
(Yes, I know, NASM 0.98.x does not support single-line macros
with arguments in %ifdef or %undef -- however, it should!  :-)

__________________________________
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/