🚀 go-pugleaf

In article <48190c$bak@nyx10.cs.du.edu>,
colin@nyx10.cs.du.edu (Colin Plumb) wrote:
>I noticed that a DEC's Alpha compiler breaks on
>
>host% cat > foo.f <<EOF
>#ifndef HAVE_FOO
>#define HAVE_FOO defined(FOO)
>#endif
>
>#if HAVE_FOO
>#error No, we have FOO
>#else
>#error Yes, we have no FOO
>#endif
>EOF
>host% cc -migrate foo.c
>cc: Warning: foo.c, line 5: Spurious token(s) ignored on preprocessor directive line.
>#if HAVE_FOO
>----^

I assume that the compiler would continue to translate the file with
the warning.  Spurious warnings are allowed in Standard C, although
the text of this one appears incorrect.

>cc: Error: foo.c, line 8: #error Yes, we have no FOO
>#error Yes, we have no FOO
>-^
>cc: Warning: foo.c, line 9: Source file does not contain any declarations.
>#endif
>------^
>
>Now, the second two diagnostics are fine, but is the compiler allowed to
>not recognize "defined(FOO)" when it appears as part of a macro expansion?

No it isn't.  Macro replacements are explicitly performed in the
constant expression following "#if ": "Prior to evaluation, macro
invocations in the list of preprocessing tokens that will become the
controlling constant expression are replaced (except for this macro
names modified by the _defined_ unuary operator), just as in normal
text".

>Now, I don't think this is allowed to fail, but my understanding of
>phases of translation is imperfect (e.g. my recent gaffe about breaking
>backslash-escapes across lines) and I'm having a hard time producing a
>derivation of this fact from the standard so I can write a sufficiently
>smug bug report.

Nothing has failed as far as the Standard is concerned (since only a
warning was given), but the warning text is bogus.

Thad