[Nasm-devel] git master branch disallows question mark in identifiers

[C. M. <pu...@38...>]

Hi,

In the commit
https://repo.or.cz/nasm.git/commitdiff/a0743398543202cc4fee33921820e33a2b5a4c3d
you changed isidstart to not allow the question mark '?' in identifiers
any longer. I understand that this is related to
https://repo.or.cz/nasm.git/commitdiff/099cc177398df447ee7153ab29337f00dbcc9d16
which adds the ternary ? : operator.

However, in my collection of macros (at
https://bitbucket.org/ecm/lmacros -- look for lmacros2.mac's lframe,
lvar, lpar, lenter, lleave) I'm using the question mark in identifiers
for stack frame labels, primarily to distinguish the labels from any
other kinds of labels. But also because I need to prefix or suffix the
labels with something to avoid existing labels interfering with %define
or %undef operation. (Ie, %undef %1 will first expand %1 to the first
macro parameter, and if that is the name of an existing define, it will
expand that define too. So %undef gets the contents of the define, not
the define name. %undef ?%1 expands the %1 but doesn't expand the define.)

A few observations:

* When using ? in the ternary operator, there must be a valid expression
term in front of it. And there must not be any unary or binary operators
there (like - and +), because these imply that the ? is the start of an
expression term itself (not the first part of the ternary operator).

* In my use case, there is never whitespace after the ?. (However, it
may be convenient when porting to "%define ? 0" for "db ?" cases.)

* In my use case, there is never a colon in the expression tail after
the question mark.

Ideally, I'd like if you could either default to or add either a switch
/ several switches (or pragmas, or whatever) to either use a heuristic,
or always treat question marks as parts of identifiers. (The latter case
might disallow using the ternary operator, which if so tips this in
favor of some heuristic.)

Regards,
ecm