You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Greg Stein <gs...@lyra.org> on 2002/02/28 02:43:01 UTC

COMPATABILITY WARNING! (was: svn commit: rev 1384 ...)

If you have existing repositories, then you will need to convert them before
using them with revision 1384 or later. If you don't, then you'll get a
mysterious "can't find repository" type of error.

See the new /trunk/tools/enable-dupkey.sh script.

Cheers,
-g

On Wed, Feb 27, 2002 at 08:21:42PM -0600, gstein@tigris.org wrote:
> Author: gstein
> Date: 2002-02-28 02:21 GMT
> New Revision: 1384
> 
> Added:
>    trunk/tools/enable-dupkeys.sh
> Modified:
>    trunk/subversion/libsvn_fs/strings-table.c
>    trunk/subversion/libsvn_fs/strings-table.h
>    trunk/subversion/tests/libsvn_fs/strings-reps-test.c
> Log:
> Switch over to the "duplicate keys" design for the 'strings' table.
> 
> COMPATABILITY NOTE: this requires an update to existing
> repositories. The strings table needs to have its "duplicates" flag
> enabled. See the 'tools/enable-dupkey.sh' script to fix up existing
> repositories.
>...
> * tools/enable-dupkey.sh: new tool to convert an existing database
>     over to the new "duplicate keys" format.
>...

-- 
Greg Stein, http://www.lyra.org/

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

Re: COMPATABILITY WARNING! (was: svn commit: rev 1384 ...)

Posted by Daniel Berlin <da...@dberlin.org>.
Doing this conversion led me to discover a bug in 
db_load, that i should warn about.

If your db_load runs forever, just consuming more and more memory, you've 
got this bug.

If your platform has char as an unsigned char by default, you have 
this problem.
You can see the real bug by compiling db_load (just go to db/dist, rm -f 
db_load*, make db_load to compile it again).
You'll see that it gives a warning about their use of "if ((ch = 
getchar()) == EOF)"

The problem is that ch is a char, rather than an int (getchar returns EOF, 
which is -1 on most platforms, on end of file), and it thus knows it'll 
never be true.

So it never thinks it sees the end of the file, and happily goes around 
and around again.

Change the declaration of ch to an int, and the problem goes away.

Compiling with -fsigned-char would also make the problem go away.

--Dan


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