#93 [PATCH] : Get autoreconf running on Ubuntu 10.10
I've attached three patches to get autoreconf on Basilisk II running without errors or warnings on Ubuntu 10.10.
First, AC_PROG_CC runs too late. If AC_PROG_CC shows up after the arch test in configure.ac, it causes something to go wrong in autoconf (at least on my box). Configure then fails to detect the .o suffix for compiled .c files correctly when AC_PROG_CC actually does run, and causes numerous other later tests involving C compilation to fail, including should-always-work ones for standard C headers.
Second, AM_PATH_GTK is undefined. AM_PATH_GTK was formerly defined in the gtk 1 development package, but Ubuntu 10.10 no longer provides a gtk 1 development header. I was a bit unsure about the right way to go about solving this; I've not much experience with automake. Finally, I decided to, on systems lacking a real AM_PATH_GTK, define AM_PATH_GTK to be a do-nothing macro that always executes as if AM_PATH_GTK had found no GTK 1 packages on the system and warn the user. This means that if a configure file is built on a system lacking the gtk 1 development headers, that configure file, when run on any subsequent system, even one with GTK 1 headers, will not be able to detect the headers. However, it also means that boxes without gtk1 can continue to run autoconf on the Basilisk II configure.ac.
Third, autoconf complains about the fact that AC_CACHE_CHECK is used improperly when testing for -fno-strict-aliasing support. This macro is apparently supposed to execute a particular command and take no actions with side-effects based on it other than setting a single variable. According to the Autoconf manual, this is a common error. This also appears to me to happen in other invocations of AC_CACHE_CHECK in configure.ac, but this is the only one that autoconf seems to complain about.
PROG_CC patch
AC_CACHE_CHECK patch
AC_CACHE_CHECK patch for Sheepshaver
Two of the same issues affect Sheepshaver; I've attached the corresponding patches.
Mistake in the gtk detection; pulling those patches until that issue is fixed.
The AM_PATH_GTK patch is unnecessary; it was created because I misunderstood how autoreconf worked. It simply needs to be specifically instructed to search the m4 directory. I will leave it out.