Re: FSFS failing regression tests?
C. Michael Pilato wrote:
>David James <firstname.lastname@example.org> writes:
>>On 8/4/05, Michael Brouwer <email@example.com> wrote:
>>>If the user enters a log message stating
>>>* foo deleted
>>>* bar modified
>>>And someone else has already deleted foo, allowing this transaction
>>>to go through will result in an incorrect log message since someone
>>>else already deleted foo. I'd say that that warrants giving the user
>>>an out of date error and letting him/her fix the log message before
>>Good argument! I'm convinced.
>>Shall we fix the test to reflect the new behaviour? fs-test looks like
>>it's still failing
>fs-test is failing for different reasons altogether, entirely
>unrelated to merges of double-deletes.
>Further, despite attempts to make corrections to folks along the way,
>I'm 100% convinced that either:
> - most of the participants in this discussion have absolutely no
> idea what is and isn't actually going on, when this double-delete
> consideration does and doesn't actually come into play, and so
> on, or
> - folks are just talking about some future behavior of Subversion
> which was not the original intent of this thread.
>And so, just in case the former is true, I repeat: the disallowing of
>double-deletes by svn_fs_commit_txn() only happens when some other
>user slips in a delete of something I wanted to delete *between the
>time I start my commit transaction and the time the pre-commit
>auto-merge code is run*.
Yup. That's why I believe that the FSFS implementation of the merge is
wrong. Granted that log messages could be made to "lie", I don't believe
svn_fs_merge correctness should be subservient to possible log message
To unsubscribe, e-mail: firstname.lastname@example.org
For additional commands, e-mail: email@example.com
Received on Tue Aug 9 23:17:35 2005
This is an archived mail posted to the Subversion Dev