issues@subversion.tigris.org writes:
> + ------- Additional Comments From kfogel@tigris.org 2002-05-02 15:17 -------
> + Wow, Hadaka. You really know how to find 'em.
> +
> + The mysterious second error is due to this code in
> + tree.c:txn_body_revisions_changed():
> +
> + /* Hack up TMP_ID so that it represents its own predecessor.
> + Node IDs come in pairs, terminated by a trailing -1. So
> + we process a pair until the ID gets down to zero, then
> + mock up the -1 so we'll process the previous one. */
> + tmp_id->digits[len - 1]--;
> + if (tmp_id->digits[len - 1] == 0)
> + tmp_id->digits[len - 2] = -1;
> +
> + Notice how if tmp_id is [0, 0, -1] -- that is, node "0.0" -- this
> + decrement algorithm fails. It produces node [0, -1], hence the
> + error :-).
> +
> + Fixing right now.
I promised Amy I wouldn't code tonight, so I'm sending this mail as a
reminder for myself (or someone else) to verify that this bug does not
also exist in svn_fs_check_related (another place that I recall where
such ID decrementing takes place).
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri May 3 02:54:56 2002