ptr & 0 not considered as constant

[nunoplopes]

Bugzilla Link	2747
Resolution	FIXED
Resolved on	Mar 12, 2010 00:57
Version	unspecified
OS	All
CC	@efriedma-quic

Extended Description

the following code is rejected by clang (with "initializer element is not a compile-time constant"):

struct sym_reg {
char nc_gpreg;
};

int sym_fw1a_scr[] = {
((int)(&((struct sym_reg *)0)->nc_gpreg)) & 0
};

[nunoplopes]

actually this happens with any arithmetic operation. e.g. the following is also rejected:
8 * ((int)(&((struct sym_reg *)0)->nc_gpreg))

[efriedma-quic]

Hmm, not sure what to do about that one... I'll have to think about it a bit.

[efriedma-quic]

Fixed in r55627.

[llvmbot]

Pointer-to-int conversions are invalid in initializers (C99 6.6P8); I don't object to this as an extension, but it should get a pedantic warning if it doesn't already.
Are there still systems around that use this construct for offsetof? That was recognized as wrong 15 years ago, I'd expect they'd all have been fixed by now.

[efriedma-quic]

Pointer-to-int conversions are invalid in initializers (C99 6.6P8); I don't
object to this as an extension, but it should get a pedantic warning if it
doesn't already.

I do have a plan for comprehensive warnings, but it requires a complete implementation of Expr::tryEvaluate.

Are there still systems around that use this construct for offsetof? That was
recognized as wrong 15 years ago, I'd expect they'd all have been fixed by now.

No, I don't think so, but there's a lot of C code floating around that uses its own macros.