From: Ken Raeburn <raeburn@mit.edu>
Date: Tue, 29 Nov 2005 07:38:45 +0000 (+0000)
Subject: Tru64 compilation fails after k5-int.h/krb5.h changes
X-Git-Tag: ms-bug-test-20060525~59
X-Git-Url: http://git.tremily.us/?a=commitdiff_plain;h=52e7daa850554e507f7df50e77041fd67ddbcca1;p=krb5.git

Tru64 compilation fails after k5-int.h/krb5.h changes

Due to some silliness in db-config.h and the Tru64 system header files, an
accidental change in the order of inclusion of certain headers caused the build
to fail.

With this patch, "make all" succeeds, but "make check" fails partway through;
I'm still investigating, and don't know at this point if there are additional
compilation problems.

* policy_db.h: Include db.h after gssrpc/types.h, to fix compilation on Tru64.

ticket: new
target_version: 1.4.4
status: open

git-svn-id: svn://anonsvn.mit.edu/krb5/trunk@17515 dc483132-0cff-0310-8789-dd5450dbe970
---

diff --git a/src/modules/kdb/db2/ChangeLog b/src/modules/kdb/db2/ChangeLog
index 0e83bcb90..fe61913f8 100644
--- a/src/modules/kdb/db2/ChangeLog
+++ b/src/modules/kdb/db2/ChangeLog
@@ -1,3 +1,8 @@
+2005-11-29  Ken Raeburn  <raeburn@mit.edu>
+
+	* policy_db.h: Include db.h after gssrpc/types.h, to fix
+	compilation on Tru64.
+
 2005-11-17  Ken Raeburn  <raeburn@mit.edu>
 
 	* policy_db.h: Include errno.h and krb5.h instead of k5-int.h.
diff --git a/src/modules/kdb/db2/policy_db.h b/src/modules/kdb/db2/policy_db.h
index c3c16b9af..8bbef800b 100644
--- a/src/modules/kdb/db2/policy_db.h
+++ b/src/modules/kdb/db2/policy_db.h
@@ -14,9 +14,22 @@
 #include <errno.h>
 #include <krb5.h>
 #include <krb5/kdb.h>
-#include <db.h>
+/* Okay, this is a bit obscure.  The libdb2 configure script doesn't
+   detect it, but on Tru64 5.1, netinet/in.h causes sys/bittypes.h to
+   be included, and that has a typedef for u_int32_t.  Because the
+   configure script doesn't detect it, it causes db-config.h to have a
+   #define for u_int32_t, so including db.h and then netinet/in.h
+   causes compilation to fail.
+
+   Since gssrpc/types.h includes netinet/in.h, including that first
+   will cause the typedef to be seen before the macro definition,
+   which still isn't quite right, but is close enough for now.
+
+   A better fix might be for db.h to include netinet/in.h if that's
+   where we find u_int32_t.  */
 #include <gssrpc/types.h>
 #include <gssrpc/xdr.h>
+#include <db.h>
 #include "adb_err.h"
 #include <com_err.h>