Sorry for the long delay in replying. I looked at this a couple of times but
still couldn't understand what is going on here.
Could anyone else take a look at this?
- Julian
S.Ramaswamy wrote:
> On 9/3/05, Julian Foad <julianfoad@btopenworld.com> wrote:
>
>>I'm not sure what it means to create a connection to a particular revision of
>>an object. In terms of generating the log, is it necessary to connect to a
>>particular revision, or will any revision work?
>>
>>If you connect to some particular revision such as the "start" or "end"
>>revision of the log, will that make the log operation more efficient than if
>>you had connected to "head" or a random revision? That's my guess of the day.
>>
>>
>>>I meant to get a ra_session to the youngest revision of the object, before
>>>using the operating revisions on that.
>>
>>Huh? The youngest revision of an object is always given by
>>svn_opt_revision_head. I suppose you meant the younger of the two specified
>>endpoints for the log, "start" and "end". Yes? But your code only does that
>>in certain cases (namely when both are specified as a number). Therefore
>>either your code is incomplete, or it is not necessary to connect to the
>>younger of the two revisions. If it is not necessary, then why do you do it in
>>some cases? Because it makes the log operation more efficient? If that's the
>>case, that's fine, but it's not obvious so a comment in the code would be
>>useful. If that's not the case then what is the reason for it?
>
>
> While it is generally true that passing start or end as the revision
> to svn_client__ra_session_from_path() doesn't change the result, there
> is one case in which it matters - if the revision range extends over
> renames, and both the revisions are numbers or dates, then the
> revision number/date passed to svn_client__ra_session_from_path()
> seems to matter.
>
> For example, if trunk/folder existed between r4 and r6 and was renamed
> to trunk/folder1 in r7, then, trying to get the log with 'svn log
> -r4:7 trunk/folder1' doesn't work if you are passing the start
> revision(4) to svn_client__ra_session_from_path(). But passing the end
> revision (7) to svn_client__ra_session_from_path() works. Here's the
> result with the actual revision # of the object(rev) and the final
> resulting url(url_p) from svn_client__ra_session_from_path() printed
> out - In this case the revision was set to start.
> -------------------------
> svn log -r4:7 trunk/folder1
> url_p:file:///home/ramaswamy/issues/2287/repos/trunk/folder
> rev:4
> subversion/libsvn_fs_fs/tree.c:315: (apr_err=160013)
> svn: File not found: revision 7, path '/trunk/folder'
> -------------------------
> Same case with dates. Setting the revision uniformly to
> svn_opt_revision_unspecified also does not work, with cases like "svn
> log -r4:6 trunk/folder1".
>
> Revised patch(v3) attached.
>
> Ramaswamy
>
> [[[
> Fix issue #2287 - add peg revision to svn_client_log2() and add
> peg revision support to the command line client.
>
> * subversion/include/svn_client.h:
> (svn_client_log3): New prototype.
> (svn_client_log2): Deprecate.
>
> * subversion/libsvn_client/log.c:
> (svn_client_log3): New function.
> (svn_client_log2): Re-implemented using new function
> svn_client_log3().
> (svn_client_log): Re-implemented using new function
> svn_client_log3().
>
> * subversion/clients/cmdline/log-cmd.c:
> (svn_cl__log): Call svn_client_log3().
>
> * subversion/tests/clients/cmdline/log_tests.py:
> (url_missing_in_head, log_through_copyfrom_history): Use peg
> revisions.
> ]]]
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Oct 10 14:59:54 2005