<p><br>
On 04/08/2011 11:47 PM, "Zooko O'Whielacronx" <<a href="mailto:zooko@zooko.com">zooko@zooko.com</a>> wrote:<br>
><br>
> > I'd suggest a kinda: if exists(darcs) ... else ... code block.<br>
><br>
> There is one. This is the way it is supposed to work. If you have<br>
> darcs in your patch then it generates the version number from darcs<br>
> history, and if it doesn't then it uses whatever version number is<br>
> left in src/allmydata/_version.py. (Brian Warner came up with this<br>
> design in about 2006.)<br>
><br>
> Therefore, people who redistribute tahoe-lafs in some container which<br>
> isn't a darcs repository should include a src/allmydata/_version.py<br>
> with it. If they don't, you'll get "version == 'unknown'". Those tests<br>
> that you saw fail are there to make sure that someone isn't running a<br>
> Tahoe-LAFS build with "version == 'unknown'". :-)<br></p>
<p>I think the problem is that when someone (i.e. me!) gets the code from the repo but doesn't have darcs. In that case would _version.py be populated?</p>
<p>I get the feeling I am just hitting a complete edge case that is _really_ rare. What are the odds of someone getting the code from the repo _without_ having darcs. Most people without darcs would prob be getting pre-packaged code.</p>
<p>I just like being difficult... :)<br><br></p>
<p>Olaf<br>
</p>