[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: switch --relocate fails on deleted dir in mixed-rev WC

From: Philip Martin <philip_at_codematters.co.uk>
Date: 2005-10-12 01:43:07 CEST

Julian Foad <julianfoad@btopenworld.com> writes:

> - if (recurse && (entry->kind == svn_node_dir))
> + if (recurse && (entry->kind == svn_node_dir)
> + && ((! entry->deleted && ! entry->absent)
> + || (entry->schedule == svn_wc_schedule_add)
> + || (entry->schedule == svn_wc_schedule_replace)))
> {
> svn_wc_adm_access_t *subdir_access;
> const char *subdir = svn_path_join (path, key, pool);
> I'm rather unsure of the exact meanings and combinations of "absent"
> and "deleted" and "schedule", and am worried by other bits of code
> using different tests such as the following:

'schedule="delete"' indicates that an existing item will be deleted at
the next commit, think 'svn rm wc/foo'.

'schedule="add"' indicates that a new item will be added at the next
commit, think 'svn add wc/foo'.

'schedule="replace"' indicates both of the above, think 'svn rm wc/foo'
followed by 'svn add wc/foo'.

'deleted="true"' indicates that the entry was scheduled for deletion
and subsequently committed, but that the parent directory has not been
updated and so is at a revision that would otherwise contain the
deleted item, think 'svn rm wc/foo' followed by 'svn ci wc'

'absent="true"' indicates that mod_authz_svn doesn't allow read access
to that item so it's not physically present in the working copy.

Your change looks correct to me.

> subversion/libsvn_wc/update_editor.c:complete_directory()
>> /* Remove any deleted or missing entries. */
>> for (hi = apr_hash_first (pool, entries); hi; hi = apr_hash_next (hi))
>> {
> [...]
>> /* Any entry still marked as deleted (and not schedule add) can now
>> be removed -- if it wasn't undeleted by the update, then it
>> shouldn't stay in the updated working set. Schedule add items
>> should remain.
>> */
>> if (current_entry->deleted)
>> {
>> if (current_entry->schedule != svn_wc_schedule_add)
>> svn_wc__entry_remove (entries, name);
> Can someone explain why that doesn't consider "replace"? If it's a
> bug, can we construct a test for it?

Not a bug. If the item is 'deleted="true"' the delete has already
been committed, so that item can't be 'schedule="delete"' as it no
longer exists, and so 'schedule="replace"' cannot happen.

That bit of code was the subject of much discussion when written and
it's still one of the hairy bits of libsvn_wc. I think it relies on
the 'incomplete="true"' processing to make it robust in the face of an
interrupted update, but I would not like to say for certain that it
always works.

Philip Martin
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Oct 12 01:44:04 2005

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.