Re: PATCH for java.nio FileChannel and MappedByteBuffer

[andreas tobler <toa at pop dot agri dot ch>]

Per Bothner wrote:
> andreas tobler wrote:
>
> (Note that the version I checked in does have this code in
> gnu/java/nio/natFileChannelPosix.cc, and uses teh symlink model.)

:) yep.

> The first question is: does it matter?  I believe for C it wouldn't,
> but C++ has I believe more restrictive rules for (void*).
>
> > What shall we do, define POSIX_C_SOURCE being bigger than 2 in this file?
> >
> > or do an ifdef for this case?
>
>
> My inclination is the former.

I had something like this in mind. Though it'll take another 24 hours to 
complete the build and test. The machine is a bit slow.

Andreas


Index: gnu/java/nio/channels/natFileChannelPosix.cc
===================================================================
RCS file: 
/cvs/gcc/gcc/libjava/gnu/java/nio/channels/natFileChannelPosix.cc,v
retrieving revision 1.1
diff -u -r1.1 natFileChannelPosix.cc
--- gnu/java/nio/channels/natFileChannelPosix.cc        29 Feb 2004 
19:09:28 -0000      1.1
+++ gnu/java/nio/channels/natFileChannelPosix.cc        8 Mar 2004 
21:11:45 -0000
@@ -55,6 +55,13 @@
 #endif

 #ifdef HAVE_MMAP
+// To make the solaris 2.6 compilation work we have to bump the version
+// of _POSIX_C_SOURCE to a value bigger than 2.
+// Defined values can be found in sys/feature_tests.h.
+# if defined (__sparc__)
+# undef  _POSIX_C_SOURCE
+# define _POSIX_C_SOURCE                 199506L
+# endif  /* __sparc__ */
 #include <sys/mman.h>
 #endif