Configured without bdb but tries to load libsvn_fs_base anyway

[Jacek Nitkiewicz <kc...@poczta.onet.pl>]

Hi All


Does it make any sense to define filesystem module "base" in fs-loader.c
if subversion is configured not to use berkeley db?

It leads to problems when you have older libs installed because e.g.
`svnadmin h' tries to load the old libsvn_fs_base and just prints:

Mismatched FS module version for 'bdb': found 1.3.0, expected 1.1.4

The following patch seems to solve the problem.  Maybe you could
consider applying it in the next version?


--- fs-loader.c.orig    2006-02-11 04:14:58.000000000 +0100
+++ fs-loader.c 2006-02-11 04:20:34.000000000 +0100
@@ -58,12 +58,12 @@
   const char *fsap_name;
   fs_init_func_t initfunc;
 } fs_modules[] = {
+#ifdef SVN_LIBSVN_FS_LINKS_FS_BASE
   {
     SVN_FS_TYPE_BDB, "base",
-#ifdef SVN_LIBSVN_FS_LINKS_FS_BASE
     svn_fs_base__init
-#endif
   },
+#endif

   {
     SVN_FS_TYPE_FSFS, "fs",



Thanks,
Jacek


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: Configured without bdb but tries to load libsvn_fs_base anyway

[Jacek Nitkiewicz <kc...@poczta.onet.pl>]

Greg Hudson wrote:
> The symbol you've conditionalized on is also undefined if Subversion is
> configured to use BDB, but is also configured with --enable-dso such
> that the module is loaded at runtime.

Oops, I overlooked it, sorry.

> We take the same approach in libsvn_ra.  We could remove the table entry
> when BDB isn't built, but it would be extra complexity.

Well, just a few additional lines to prevent the confusing behaviour, at
least for libsvn_fs_base (I saw similar problems on users@ list:
http://svn.haxx.se/users/archive-2005-08/0609.shtml).

Could you please have a look at the attached patch?


Regards,
Jacek

Re: Configured without bdb but tries to load libsvn_fs_base anyway

[Greg Hudson <gh...@MIT.EDU>]

On Sat, 2006-02-11 at 06:28 +0100, Jacek Nitkiewicz wrote:
> The following patch seems to solve the problem.  Maybe you could
> consider applying it in the next version?

The symbol you've conditionalized on is also undefined if Subversion is
configured to use BDB, but is also configured with --enable-dso such
that the module is loaded at runtime.

We take the same approach in libsvn_ra.  We could remove the table entry
when BDB isn't built, but it would be extra complexity.


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org