RE: test for restrict fails with MS compiler

[Jerker Bäck]

Hello Ralf,
I seem to have a habit of forgetting to add the group

> <http://thread.gmane.org/gmane.comp.sysutils.autoconf.general/7628>.
Ah, I see where you got this from.

Well, maybe you're right. I will be using the MS implementation of restrict
and see if I run into problems. There could be a risk of wrong optimization
- we'll see.

> Be encouraged to post about the issues you have.  AFAIK the w32 platform
> does not have any mmap support whatsoever, no?  And as to stdbool, I
> guess I should recommend the stdbool module of gnulib, which is also
> mentioned in Autoconf's manual (where AC_HEADER_STDBOOL is discussed).

There are memory mapped files ("File mapping"), in fact almost all NT kernel
and userspace core libraries file I/O are carried out this way. One common
developer technique is to use the system paging file for temporary files
(CreateFileMapping is called with INVALID_HANDLE_VALUE). This file handle
can then be used as a buffer in for example texteditors without the need to
allocate memory for the text entered or pasted. But most of these functions
are Win32 specific. In POSIX we have the feature implemented in another way:

Interix SDK prototype:
void * __cdecl mmap (void *, size_t, int, int, int, off_t);

The autoconf mmap test fails. The reason for this is complicated and hard to
explain, but is due to the tests requirement for fixed memory (MAP_FIXED).
In short: It violates the whole idea with the NT kernel Virtual Memory
Manager. You know, there are no real memory in Windows, only virtual.

Cheers, Jerker