Re: svn_client_status5() and depth

On 17.08.2010 23:41, Hyrum K. Wright wrote:
> On Tue, Aug 17, 2010 at 3:45 PM, C. Michael Pilato<cmpilato_at_collab.net> wrote:
>> On 08/17/2010 01:42 PM, Stefan Küng wrote:
>>> Maybe I don't understand that change:
>>> --depth specifies a depth to use for the command. If I want the command
>>> to use the depth of the working copy, I specify an unknown depth or none
>>> at all. But if I specify a depth, I would assume the command to respect
>>> that depth and return the info with that depth.
>>> So why should the -u flag not use the specified depth?
>>
>> --depth is a filtering option -- it can only reduce the scope of an
>> operation, it can not expand it.
>>
>> Bert sez he'll make the behavior optional in the API call so you can make
>> use of it. We won't expose it at all in the 'svn' command-line client, but
>> TortoiseSVN can get its behavior back. Sound good?
>
> <sarcasm>
> Yay for another API flag!
> </sarcasm>

Well, it's either another flag, or use bitmasks. And since bitmasks are
not allowed anymore, there aren't any more options.

> I'm starting to wonder if the mission (at the API level) of 'st -u'
> and vanilla 'st' has diverged enough that we may want to consider an
> API named svn_client_remote_status() or something. It seems a bit
> silly to keep extending APIs using various boolean flags[1], in a
> manner which eventually leads to both consumer and programmer
> confusion. This eventually leads to entry points whose behavior
> wildly varies based upon the permutations of the set of flags.
> Perhaps status5 (with its 13 parameters) has reached this point?

But that would lead to another problem: how would you combine the local
status and the remote status data? Have the clients merge those two?

Bert:
I've checked r986510 and I like to add some comments:

In libsvn_client/deprecated.c, svn_client_status4() you pass FALSE for
the new sticky param. But that's wrong: in 1.6.x and previous versions,
the depth was always respected/enforced. That's why I noticed the
changed behavior in the first place!
Shouldn't you pass TRUE there?

And what I don't understand: why the extra parameter at all?
If the param is set to TRUE unconditionally, then the behavior seems to
be the very same as in 1.6.x:
If the depth is less than the depth of the WC, it behaves the same as if
the param is FALSE. Only if the depth is deeper then the param has any
effect - setting the param to TRUE *ignores* deeper depths.

So from my point of view, that parameter isn't necessary: the behavior
can be configured by using the depth parameter alone.

Also, since this is a readonly operation, the naming 'depth_is_sticky'
isn't very good. Even confusing.

Ok, reading my text above again, I think I'm not very clear why I think
the parameter isn't necessary. So I'll try again:

depth sticky
<= WC depth FALSE 1.6.x == 1.7 : data for depth
<= WC depth TRUE 1.6.x == 1.7 : data for depth
> WC depth FALSE 1.6.x data for depth, 1.7 data for WCdepth
> WC depth TRUE 1.6.x data for depth, 1.7 data for depth

Also, if a client wants the data only for the WC depth, passing a depth
of 'unknown' will do exactly that, no matter what the sticky is set to.

So: every outcome of that API can be triggered by specifying the depth
alone. So why not get rid of the sticky param and always do as if it was
set to TRUE?

Stefan

-- 
        ___
   oo  // \\      "De Chelonian Mobile"
  (_,\/ \_/ \     TortoiseSVN
    \ \_/_\_/>    The coolest Interface to (Sub)Version Control
    /_/   \_\     http://tortoisesvn.net