Re: [Issue 2419] New - 'svn rm' of schedule add leaves working props file

Erik Huelsmann <ehuels@gmail.com> writes:

>> ------- Additional comments from philip@tigris.org Mon Oct 17 17:00:23 -0700 2005 -------
>>
>> rm and revert on a schedule add file should be similar, the only difference
>> being that rm deletes the working file while revert leaves it unversioned, yet
>> the implementations are totally different: rm uses log files while revert
>> modifes the working copy directly. Does anyone know how these things are
>> supposed to work?
>
> Yes, I do.
>
> The effects should be equal and both svn_wc_delete2 and svn_wc_revert2
> should be using logs. Actually, more generally: every working copy
> action which modifies more than 1 state in an operation
> (svn_wc_delete2 and _revert2 both change working copy files, props
> files and entries, which add up to a minimum of 3 changes) should use
> log files to do the work.

Using log files can work, but it's not always necessary or sufficient.

It's not necessary when all we need to do is ensure that the wc
remains in a consistent state if interrupted. If the direct actions
do that then log files are unnecessary.

It's not sufficient since simply using log files doesn't guarantee
that the operation can be repeated, see issue 2424.

> The conclusion should be we need another issue to record the fact that
> svn_wc_delete2() isn't loggy.

Delete is loggy, perhaps you meant revert?

> Does this answer your question, or did you mean to ask a different one?

Heh. I was really complaining about the state of the wc code, it's no
fun fixing bugs when two such similar operations have completely
different implementations. I might well be responsible for some of
the mess :-(

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