Blair Zajac wrote:
> Ben Collins-Sussman wrote:
>> On Sat, Sep 27, 2008 at 7:49 AM, Branko Čibej <brane_at_xbc.nu> wrote:
>>> Barry Scott wrote:
>>>> I'm fixing a problem in pysvn Workbench that is a result of
>>>> svn_client_logX returning
>>>> revision 0.
>>>>
>>>> Revision 0 is not like the other revisions because it only the
>>>> svn:date
>>>> revprop.
>>>>
>>>> The svn log command filters out any entry that has a revision of 0
>>>> and any
>>>> that
>>>> do not have a svn:message. (Filtering out svn:message == NULL seems
>>>> very
>>>> odd
>>>> isn't svn:message mandatory in a repos?).
>>> What is svn:message?
>>>
>>> And WFIW, no revprops should be mandatory, ever.
>>>
>>
>> No revprops are ever required to exist. No 3rd-party program should
>> ever assume that 'svn:date', 'svn:author', or 'svn:log' exist; it's
>> not an error for any of them to not exist. svn:author only exists if
>> the commit was authenticated; svn:log only exists if someone typed a
>> message; and even though svn:date is added by the repository at
>> commit-time, a committer might have used 'svn propchange' to simply
>> remove it. These are all legitimate situations.
>
> Why would we ever allow svn:date to be removed?
Because it isn't required for proper operation? Sure, it makes
date->revision conversions somewhat ... impossible. But Subversion's
primary -- only required -- index for changes is its revision numbers.
--
C. Michael Pilato <cmpilato_at_collab.net>
CollabNet <> www.collab.net <> Distributed Development On Demand
Received on 2008-09-27 22:15:47 CEST