> Actually, the real problem is that the Cygwin-targeted version of gnat
> shouldn't need those definitions in the first place. Cygwin provides a
> fairly complete Linux/Posix feature-set, and doing an end-run around it by
> using the underlying winsock API isn't usually a good idea, so I think that
> the better solution is to switch over to the standard berksock
> implementation.
This sounds fine.
> * gcc-interface/Makefile.in (WIN_TARG_SUFFIX): New variable, used by
> windows targets only to differentiate between MinGW and Cygwin.
> (LIBGNAT_TARGET_PAIRS [windows targets]): Correctly detect cygwin,
> which no longer has the '32' suffix, and use WIN_TARG_SUFFIX to choose
> appropriate implementations of the sockets and memory packages.
Ouch. :-) Please do something like this:
# On Cygwin, we use the default version of s-memory and g-socthi because
# <insert some (sketch of a) reason here>.
ifeq ($(strip $(filter-out cygwin%,$(osys))),)
LIBGNAT_TARGET_PAIRS = \
s-memory.adb<s-memory.adb \
g-socthi.ads<g-socthi.ads \
g-socthi.adb<g-socthi.adb
else
LIBGNAT_TARGET_PAIRS = \
s-memory.adb<s-memory-mingw.adb \
g-socthi.ads<g-socthi-mingw.ads \
g-socthi.adb<g-socthi-mingw.adb
endif
LIBGNAT_TARGET_PAIRS += \
a-dirval.adb<a-dirval-mingw.adb \
[...]
g-soliop.ads<g-soliop-mingw.ads \
$(ATOMICS_TARGET_PAIRS)
> * sysdep.c (WIN_SETMODE): New define to choose the correct spelling of
> setmode/_setmode for MinGW and Cygwin, respectively.
> (__gnat_set_binary_mode [windows targets]): Use the above, and enable
> the windows version for Cygwin as well as MinGW.
> (__gnat_set_text_mode [windows targets]): Likewise.
> (__gnat_ttyname [windows targets]): Provide a Cygwin implementation
> in addition to the MinGW version.
> (__gnat_is_windows_xp): Make available to Cygwin as well as MinGW.
> (__gnat_get_stack_bounds): Likewise.
OK with the above change, thanks for fixing this.
--
Eric Botcazou
