MMAP issue on cygwin

[Zack Weinberg]

On Tue, Jan 01, 2002 at 03:36:05PM -0500, Christopher Faylor wrote:
> On Tue, Jan 01, 2002 at 12:18:43PM -0800, Zack Weinberg wrote:
> >I think the proper resolution to this issue is to chuck out the
> >elaborate tests in aclocal.m4 that try to pin down whether or not mmap
> >works.  Instead, check whether the function exists at all, and then
> >blacklist systems known not to work.  I believe that'd be just cygwin.
> 
> It's not just "cygwin".  It's "cygwin on Windows 95/98/Me".  It should
> work ok on Windows NT.

Is it practical to tell the difference at configure or run time?

> IIRC, reverting to non-mmmaped cases caused a
> severe performance penalty so the decision to avoid this should not be
> taken lightly.

That was the garbage collector's use of mmap to allocate memory.
cpplib uses it to read input files, which is a much less important
perfomance optimization.

> We've been torturously following along as you implement more stringent
> mmap requirements.  We're looking into fixing this for 95 now, just
> like we fixed every other mmap problem that gcc unearthed.

We appreciate the work you've done on this.

zw