Suggested changes

[ra...@madhaus.utcs.utoronto.ca]

It is a bit irritating that the Apache conf.h file uses HAVE_MMAP to
determine whether it should include sys/mman.h.  HAVE_* is the standard
naming mechanism for autoconf-generated configure scripts and in the case
of the mmap() call, there are systems out there that have mmap but do
not have the sys/mman.h header file.  This means that any Apache module
that uses mmap and uses configure to check for the presence of this function
will have compile problems on certain platforms.  I think that if we are
not going to use autoconf for Apache, then we could at least make sure that
the Apache-specific macros do not conflict with autoconf.

Also, how about removing the debug code from alloc.c?  I think the memory
pool code has been tested now.  The calls to chk_on_blk_list() and the
function itself can probably be removed.  The memset() call in make_sub_pool()
could probably go too.  Is there a good reason to 0 out the pool on creation?

-Rasmus

Re: Suggested changes

[Tom Tromey <tr...@creche.cygnus.com>]

Rasmus> It is a bit irritating that the Apache conf.h file uses
Rasmus> HAVE_MMAP to determine whether it should include sys/mman.h.
Rasmus> HAVE_* is the standard naming mechanism for autoconf-generated
Rasmus> configure scripts and in the case of the mmap() call, there
Rasmus> are systems out there that have mmap but do not have the
Rasmus> sys/mman.h header file.

As part of my "autoconfiscation" of Apache, I changed all the internal
macro names to conform to Autoconf's expectations.  I imagine I'll do
this again when I move to 1.1.  I'm still not sure exactly when that
is going to happen.  Hopefully by the end of this month, but (as usual
:-) things are a little confused right now.

Tom
-- 
tromey@cygnus.com                 Member, League for Programming Freedom