Everhart,Glenn
From:	John D. Amidon [john.d.amidon@cpmx.saic.com]
Sent:	Sunday, April 19, 1998 6:15 PM
To:	Info-VAX@Mvb.Saic.Com
Subject:	Re: CVS for VMS 7.1 Alpha?
>>>>> "Jim" == Jim Kingdon <kingdon@panix3.panix.com> writes:

>> I had CVS working on a VMS 6.2 Alpha system, but we had to upgrade to VMS
>> 7.1.  Now it crashes in . . . getpass ["cvs login" command]

Jim> I've seen this too.

Jim> I'm still not sure how to fix it.  CVS is trying to use curses to read
Jim> a password without echoing it to the screen, but if SMG$ or a low-level
Jim> call (RMS or SYS$QIO or whatever) to turn off echoing would be a better
Jim> way, I'd consider that.

Jim> Building under VMS 7.1 might help too, I don't know.  There are
Jim> annoying problems with include files (which I suppose I should fix
Jim> anyway).

We have built the latest 1.9.26 on OpenVMS 7.1 and will be submitting the
patches shortly.  The `login' problem is gone.  However, we are still
looking at the `release' command which does not work.

Basically there are two problems with compiling on OpenVMS 7.1, choosing the
correct 'C' standard and the misuse of `lstat' rather than the macro
`CVS_LSTAT'.

These are hacks but small ones ;) If these patches are applied correctly,
CVS should build without error on OpenVMS.

The first one was solved by applying the following patch to `vms.h':

===================================================================
RCS file: /export/repository/gnu/cvs/vms/vms.h,v
retrieving revision 1.1.1.1
retrieving revision 1.2
diff -p -u -r1.1.1.1 -r1.2
--- gnu/cvs/vms/vms.h	1998/04/04 06:04:24	1.1.1.1
+++ /export/repository/gnu/cvs/vms/vms.h	1998/04/04 11:53:53	1.2
@@ -1,6 +1,8 @@
 /* Determined from CC RTL function prototypes in online documentation */
 
-#define mode_t unsigned int
+#if !defined(__CVS_VMS_H__)
+#define __CVS_VMS_H__
+
 
 #define fork(x) vfork(x)
 
@@ -26,9 +28,17 @@ extern int fnmatch(char *pattern, char *
 #include "pwd.h"
 #include "pipe.h"
 
+#if !defined(_DECC_V4_SOURCE)
 int unlink(char *path);
 int link(char *from, char *to);
+#else
+typedef __off_t off_t;
+typedef __ino_t ino_t;
+#define mode_t unsigned int
+#endif
 
 #define stat(a, b) wrapped_stat(a, b)
 
 #undef POSIX
+
+#endif

and by adding the definition of `_DECC_V4_SOURCE' to the `/define'
statements in the build_diff.com, build_src.com, build_vms.com,
build_zlib.com, and build_lib.com.  In addition, also add `__CVS_VMS_H__' to
the '/define' statement for build_lib.com.

The second was fixed by applying the following patch to `stat.c' in the
`vms' directory:

===================================================================
RCS file: /export/repository/gnu/cvs/vms/stat.c,v
retrieving revision 1.1.1.1
retrieving revision 1.2
diff -p -u -r1.1.1.1 -r1.2
--- gnu/cvs/vms/stat.c	1998/04/04 06:04:24	1.1.1.1
+++ /export/repository/gnu/cvs/vms/stat.c	1998/04/04 11:53:53	1.2
@@ -31,3 +31,14 @@ struct stat *buffer;
 
   return rs;
 }
+
+/*
+ * lstat is non-existent on VMS
+ * fake here so we don't have to change other sources
+ */
+int lstat (path, buffer)
+const char *path;
+struct stat *buffer;
+{
+  return wrapped_stat (path, buffer);
+}

However, I believe if everybody followed the rules and used the macros where
they are suppose to this patch would not be needed.

Let me know if I can be of further assistance.
-- 
---------------------------------------------------------------------------
John D. Amidon, TransCore, 3500 Parkway Lane, Suite 600, Norcross, GA 30092
e-mail: john.d.amidon@cpmx.saic.com | V:+1 770 447 6831 | D:+1 770 449 7268
---------------------------------------------------------------------------