Bug 21080

My munmap man page says
 int munmap(void *addr, size_t length);

And sun docs agree http://docs.sun.com:80/app/docs/doc/816-5167/munmap-2?a=view

Hmm, it would appear that the definition changes depending on some #defines... 

From /usr/include/sys/mman.h:

#if (_POSIX_C_SOURCE > 2) || defined(_XPG4_2)
...
extern int munmap(void *, size_t);
...
#else	/* (_POSIX_C_SOURCE > 2) || defined(_XPG4_2) */
...
extern int munmap(caddr_t, size_t);
...

It looks like for C++ these macros are not defined since they conflict with the ANSI C++ standards - from standards manpage:

  ANSI/ISO C++
     ANSI/ISO C++ does not define any feature test macros. If the
     standard C++ announcement macro __cplusplus is predefined to
     value  199711  or	greater,  the  compiler	 operates  in	a
     standard-conforming  mode,	 indicating C++	standards confor-
     mance. The	value 199711  indicates	 conformance  to  ISO/IEC
     14882:1998,  as required by that standard.	 (As noted above,
     conformance to the	standard  is  incomplete.)   A	standard-
     conforming	mode is	not available with compilers prior to Sun
     WorkShop C++ 5.0.


     C++ bindings are not defined for POSIX  or	 X/Open	 CAE,  so
     specifying	  feature  test	 macros	 such  as  _POSIX_SOURCE,
     _POSIX_C_SOURCE, and _XOPEN_SOURCE	can result in compilation
     errors  due  to conflicting requirements of standard C++ and
     those specifications.

It would appear to not be a bug after all, I just need to figure out the correct definitions for Solaris.

Apologies, closing as not a bug.

Created attachment 24664 [details] [review]
Try again with #ifdefs...

I've changed my mind, this is still an issue - since defining any of the "standards" macros causes more serious breakages else where. 

As a compromise, I've put a specific #ifdef test around the munmap call to meet the Sun Compiler's settings, so it should have no effect else where.

Reopening since no mix of #defs allows all of poppler to compile cleanly. It makes more sense to fix specifically for the Sun Compiler.

Carlos, what you think of this?

if data->bytes is an unsigned char * and munmap expects a char * why not just adding the cast without any defines?

Because of comment #1

hmm, then we should use (void *) instead of (char *), no?

But it seems some Solaris version and some Solaris compiler doesn't like void*

strange. Well, if the patch fixes the problem and doesn't break any other compilers/platforms, it's ok to me. 

Fixed in git