Andrew Pinski wrote:
>
> On Nov 16, 2004, at 1:30 PM, Dave Hines wrote:
>
> > Possible bug in gcc-3.4.3 fixincludes on IBM AIX 5.2
> >
> > Building the compiler ("make bootstrap") fails when compiling
> > genautomata.c, due to a header file problem. The header file which
> > produces the error is generated by fixincludes from the system math.h
> > file. It appears that the word "class" is triggering the insertion of
> > an #ifndef...#endif pair, but the #endif is being tacked on to the end
> > of the preceeding line, which ends in a continuation character.
>
> This was PR 11315 but it is a bug in AIX really.
>
> Quote from the bug:
>
> While GCC fixincludes could work around this, it really is an AIX
> problem. This has been opened as a defect against AIX 5.2 and will
> be fixed by IBM APAR IY46553. It should be avaliable from IBM
> Customer Support website in a few weeks (I'll add the PTF number
> when I find out) and should appear in the next maintenance release.
That has to be a very interesting bug! Here is the fix specification,
with a translation below it:
> fix = {
> hackname = rs6000_double;
> files = math.h;
> select = '[^a-zA-Z_]class\(';
>
> c_fix = format;
> c_fix_arg = "#ifndef __cplusplus\n%0\n#endif";
> c_fix_arg = '^.*[^a-zA-Z_]class\(.*';
>
> test_text = "extern int class();";
> };
The regex library is used to search for a line that matches the
pattern `^.*[^a-zA-Z_]class\(.*' [[A line that contains the word
"class" before an opening parenthesis that is not the end of a longer
token.]] The fixincl program itself then inserts:
"#ifndef __cplusplus\n"
before the matched text and:
"\n#endif"
after it. How can the trailing "#endif" get attached to the
preceding line? Some weird I/O bug? Oh, well.....Not a lot
I can do about it (add a second newline?)....Sorry - Bruce
