Re: Consistency between 'update' and 'switch' APIs

Julian Foad <julianfoad_at_btopenworld.com> writes:

> Philip Martin wrote:
>> Michael Pilato" <cmpilato_at_collab.net> writes:
>
>>
>>> [Deferring further comment until we unearth the reasoning behind switch
>>> supporting --ignore-ancestry at all.]
>>
>> You added it in r1087015 as a backwards compatiblity switch when fixing
>> issue http://subversion.tigris.org/issues/show_bug.cgi?id=3848.
>
> IMO what we need to get to the bottom of, is not the command-line
> switch '--ignore-ancestry', but rather the reason why the switch
> command ever requests ignore-ancestry behaviour of the repository-side
> diff that it performs.

Backwards compatibility. svn_repos_dir_delta had ignore-ancestry since
1.0, it was added in r845731, and switch operated with ignore-ancestry
set false. r1087015 started making switch require history between the
nodes, with a switch to get the old behaviour. At this stage the
history check was purely a client side check of the relationship between
the nodes, the RA/repos layers still ignored ancestry during switch.

r1449413 passes the ignore-ancestry flag over RA to repos so that switch
now takes full account of ancestry.

-- 
Certified & Supported Apache Subversion Downloads:
http://www.wandisco.com/subversion/download