Re: Broken autoconf mmap test

[Bob Friesenhahn]

On Tue, 22 Mar 2011, Ken Brown wrote:

> For a long time, the autoconf mmap test has failed on Cygwin even though 
> Cygwin has a working mmap.  This was reported to bug-autoconf in November 
> 2009, and a patch that fixed it for Cygwin was submitted by Eric Blake:
>
>  http://lists.gnu.org/archive/html/bug-autoconf/2009-11/msg00015.html
>
> It was reported in
>
>  http://lists.gnu.org/archive/html/autoconf-patches/2009-11/msg00017.html
>
> that the patch didn't help on HP/UX 11, which is another platform where the 
> test was failing.  I can't find any follow-up to this in the archives of the 
> autoconf mailing lists.

The original intent of this macro was to only pass if MMAP I/O and 
file I/O are coherent.  In other words, programmed file I/O is 
immediately reflected in already memory mapped file data (and 
vice-versa). This is not the case for Windows and HP-UX.  Given this, 
Eric Blake's subsequent response to your email leaves me concerned.

In GraphicsMagick I use a somewhat modified version to test if mmap 
works without requiring coherency.

Bob

> Is there any chance of getting this bug fixed?  Currently, the cygport tool 
> (which many Cygwin package maintainers use to build packages) works around it 
> by exporting
>
>  ac_cv_func_mmap_fixed_mapped=yes
>
> in the build script.  But it would be better to get a solution in autoconf 
> itself.
>
> Ken
>
> _______________________________________________
> Autoconf mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/autoconf

--
Bob Friesenhahn
address@hidden, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/