[Monotone-devel] 0.26pre2's configure: problems on cygwin

[Lapo Luchini]

It's been a while after I last were able to compile monotone on
cygwin, but being the error full of STL stuff and not having much time
to watch it closely, and having recently had some HDD corruption, I
put that apart with something like "must be my STL with some files
corrupted".
But today I checked closer.
The real "problem" seem to be that between 0.26pre1 and 0.26pre2
something happened that changed configure: it now tries to compile
win32/read_password.cc instead of unix/read_password.cc and, in doing
so, it exits with lots of errors (I guess win32/ directory contains
VC-only things? or mingw-only things? anyway, with cygwin's gcc, it is
no good).

Diff between the two configure outputs is, simply:
-checking if this is Windows... no
+checking if this is Windows... yes

Relevant diff in configure.ac seems to be:
--- ../monotone-0.26pre1/configure.ac   2006-01-10 02:48:52.000000000
+0100
+++ configure.ac        2006-02-11 22:40:51.000000000 +0100
@@ -290,20 +277,11 @@

 AC_CACHE_CHECK([if this is Windows],
        ac_win32, [
-       AC_TRY_RUN([
-#ifdef WIN32
-#include <windows.h>
-int main(void)
-{
-  return 0;
-}
-#else
-__wont_compile_
-#endif
-],
+       AC_TRY_COMPILE([#include <windows.h>],
+                      [HANDLE h; DWORD d;],
         ac_win32=yes,
-        ac_win32=no,
-        ac_win32=no)])
+                      ac_win32=no)
+])
 if test "$ac_win32" = "yes"; then
    AM_CONDITIONAL(WIN32_PLATFORM, true)
 else

I can confirm yesterday's trunk to compile (and work!), once that test
is reverted to the old state.

Wow "monotone-0.26pre2 annotate configure.ac" seems to be a rather
longish operation 0=)
Relevant output 15 minutes later:

077c3a5a.. by graydon
ce85ac6d.. by kinetik
c62eb45e.. by kinetik
f50199b0.. by graydon

    Lapo