As we've just found out on IRC, Zack's test failures is related to
the fact that apr-util wasn't told where Berkeley DB was. So,
apr-util was snarfing in db2 which blew the build up.
As we know, apr-util and SVN have to be compiled against the same
BDB or bad things will happen.
When using apr-util source directories (if you have a pre-built
apr-util, you probably aren't hit by this):
1) User has Berkeley DB 4 on the system. All is fine. SVN and
apr-util will pick it up since they look in relatively the
same places.
2) User has Berkeley DB 4 installed somewhere else, but passes
--with-berkeley-db to SVN's configure. This option is passed
to apr-util's configure. All is fine.
3) User doesn't have BDB installed, and places the source in ./db.
No --with-berkeley-db option is passed. SVN has special logic
for handling DB4's source location. However, apr-util doesn't
exhibit this same intelligence (and I almost don't want to add
it). I'm thinking we could pass --with-berkeley-db=no to
apr-util's configure (see below) to prevent apr-util from
recognizing any other Berkeley DB in this case.
Thoughts? -- justin
Index: ./ac-helpers/aprutil.m4
===================================================================
--- ./ac-helpers/aprutil.m4
+++ ./ac-helpers/aprutil.m4 Thu Feb 14 20:37:04 2002
@@ -22,7 +22,11 @@
fi
if test $apu_found = "reconfig"; then
- SVN_SUBDIR_CONFIG(apr-util, --with-apr=../apr)
+ APR_UTIL_OPTIONS="--with-apr=../apr"
+ if test -d "./db"; then
+ APR_UTIL_OPTIONS="$APR_UTIL_OPTIONS --with-berkeley-db=no"
+ fi
+ SVN_SUBDIR_CONFIG(apr-util, $APR_UTIL_OPTIONS)
SVN_SUBDIRS="$SVN_SUBDIRS apr-util"
fi
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:37:08 2006