Re: Windows build still requires apr-util?

I've been using VC6 to test the ruby bindings as I haven't yet succeeded in
building with mingw.

It needs this patch:

[[[
Index: build/generator/gen_win.py
===================================================================
--- build/generator/gen_win.py (revision 1454760)
+++ build/generator/gen_win.py (working copy)
@@ -401,7 +401,6 @@
       install_targets = [x for x in install_targets
                                      if not (isinstance(x,
gen_base.TargetJava)
                                              or isinstance(x,
gen_base.TargetJavaHeaders)
- or isinstance(x,
gen_base.TargetSWIGProject)
                                              or x.name == '__JAVAHL__'
                                              or x.name ==
'__JAVAHL_TESTS__'
                                              or x.name == 'libsvnjavahl')]
Index: subversion/include/svn_types.h
===================================================================
--- subversion/include/svn_types.h (revision 1454760)
+++ subversion/include/svn_types.h (working copy)
@@ -29,6 +29,7 @@

 /* ### this should go away, but it causes too much breakage right now */
 #include <stdlib.h>
+#include <limits.h> /* for ULONG_MAX */

 #include <apr.h> /* for apr_size_t, apr_int64_t, ... */
 #include <apr_errno.h> /* for apr_status_t */
Index: subversion/libsvn_subr/sysinfo.c
===================================================================
--- subversion/libsvn_subr/sysinfo.c (revision 1454760)
+++ subversion/libsvn_subr/sysinfo.c (working copy)
@@ -28,8 +28,14 @@
 #define PSAPI_VERSION 1
 #include <windows.h>
 #include <psapi.h>
+#if defined(_MSC_VER)
+#if _MSC_VER > 1200
 #include <Ws2tcpip.h>
 #endif
+#else
+#include <Ws2tcpip.h>
+#endif
+#endif

 #define APR_WANT_STRFUNC
 #include <apr_want.h>
Index: subversion/libsvn_subr/cache-membuffer.c
===================================================================
--- subversion/libsvn_subr/cache-membuffer.c (revision 1454760)
+++ subversion/libsvn_subr/cache-membuffer.c (working copy)
@@ -116,11 +116,11 @@
 /* By default, don't create cache segments smaller than this value unless
  * the total cache size itself is smaller.
  */
-#define DEFAULT_MIN_SEGMENT_SIZE 0x2000000ull
+#define DEFAULT_MIN_SEGMENT_SIZE 0x2000000ui64

 /* The minimum segment size we will allow for multi-segmented caches
  */
-#define MIN_SEGMENT_SIZE 0x10000ull
+#define MIN_SEGMENT_SIZE 0x10000ui64

 /* The maximum number of segments allowed. Larger numbers reduce the size
  * of each segment, in turn reducing the max size of a cachable item.
@@ -133,7 +133,7 @@
 /* As of today, APR won't allocate chunks of 4GB or more. So, limit the
  * segment size to slightly below that.
  */
-#define MAX_SEGMENT_SIZE 0xffff0000ull
+#define MAX_SEGMENT_SIZE 0xffff0000ui64

 /* We don't mark the initialization status for every group but initialize
  * a number of groups at once. That will allow for a very small init flags
]]]

Though honestly, I haven't much interest in it once I can get a mingw
going....

--
Joe
On Wed, Mar 13, 2013 at 10:14 AM, Ben Reser <ben_at_reser.org> wrote:
> On Wed, Mar 13, 2013 at 6:59 AM, Branko Čibej <brane_at_wandisco.com> wrote:
> > Nor should you even try to, IMO. There comes a time when it ceases to
> > make any sense to support an old compiler that's been EOL-d a while ago,
> > especially since free versions of newer compilers are available.
> >
> > I would support removing the .dsp generator completely.
>
> +1, it's old and broke and better to not exist.  If it's there someone
> is likely to try to use it.
>