C99 allows for the Single Unix Spec semantics which prescribe a return
value of -1 for n == 0 which is treated as an encoding error.
This reenables use of snprintf on Solaris at least.
Relevant excerpts from ection 7.19.6.5:
Description
[..] If n is zero, nothing is written, and s may be a null pointer. [..]
Returns
The snprintf function returns the number of characters that would have been
written had n been sufficiently large, not counting the terminating null
character, or a negative value if an encoding error occurred.
Single Unix Spec:
http://www.opengroup.org/onlinepubs/007908799/xsh/fprintf.html
The GNU people strike again with their useless docs; fopencookie does NOT
use "fpos_t *" as documented :/
Additionally, it appears that the the libio layer buffers in chunks of 8KB,
so our test needs to exploit this in order to function correctly.
Thanks to Derick for doing some groundwork.
which is 'executed' last of the SAPI config.m4's.
- Added --disable-cgi option. (was possible only with above change)
- Made the sed check to not test for the possibly working sed on Solaris
since that would make the test quite useless. (compile would still fail)
# That sed check is not enabled yet..need to know whether we fail
# during configure or just put out a warning about possible non-working
# sed. As even plain ./configure doesn't work in most cases, I think
# it would be better to just fail during configure and let the users
# fix their paths/install GNU sed.
very long strings.
#
# Sascha, is this okay? I added this here since any libtool/autoconf
# release out there doesn't have this yet..we can remove this when
# we can really start requiring such versions which have it?
#
# This sets $SED to the correct binary, so that should be used in
# places were the lines might be very long.
#
The bug causes the kernel not to return -1/EAGAIN. The new test case
has been borrowed from the Linux Test Project.
This also fixes a bug which apparently caused HAVE_PREAD/WRITE to be
defined even if the more complex checks failed (ac_cv_func_NAME=no
was set albeit with no difference).
The old checks supposed that pread/pwrite worked, if a declaration was
found in <unistd.h>. We now actually check whether they work successfully
before using them.
uses off_t or the newer, more portable "fpos_t *".
The check could perhaps be more refined, as the test program will segfault
on older systems (like mine) that use off_t.
func and __func using LIBS, and if unsuccessful, searches
each specified lib. Defines HAVE_FUNC, if found. If func/__func is
found in library foo, HAVE_LIBFOO is defined.
(Jani, me)
2. Autoconf 2.5x is more pedantic regarding locating header files.
We include the proper header files now in the check. (me)
3. The nsl/bind/socket/etc checks have been rewritten using PHP_CHECK_FUNC.
This ensures that no extra library is used, if the symbol is
available in libc which avoids issues on BSD/OS, OpenBSD and others.
(Jani)
Sascha Schumann