Re: rcs tags and tests

Dumas Patrice wrote:

>>Maybe.  I just really dislike introducing CVS commands into my build 
>>scripts.  It means that if you break a test run in the middle your 
>>sandbox will be corrupt.  That or you keep a second sandbox you use to 
>>    
>>
>
>I don't really understand that point. Are you saying that in this case
>the test.sh script would do a checkout with the right options before doing
>the diffs ?
>  
>

That's one of the ways to accomplish this.  I suppose a sed script would 
be another.  And the cvs admin -k* options being another, but I'll deal 
with that in a second.

>>run tests.  If you really don't mind having to check in anything you 
>>want to test and checking it out again in the new sandbox, maybe it's 
>>not so bad, but a programmatic solution seems much less cumbersome in 
>>the long run.
>>    
>>
>
>And it would work for anyone.
>

Exactly.

>>I guess you could use `cvs admin' to give all the relevant files -ko 
>>attributes in the repository too.  Then only they would be affected and 
>>any new sandbox checkout from the central repository would be ready to 
>>go, but you still can take a tarball distribution and check it into a 
>>new CVS repository and have it work, which is usually one of my design 
>>goals.
>>    
>>
>
>This solution seems to be the best. I have seen in the texinfo manual
>that there is a -k* option for cvs admin which does exactly that.
>  
>

Yes.  This is my second choice, but I would prefer something 
programmatic.  If someone were to tar the distribution and check it into 
another CVS repository, the tests would still break since their CVS 
repository would replace the keywords on checkout.   This could 
discourage some novice developers.  Maybe I'll try and make some time to 
add the @ifhtml stuff sometime this week or next.

Derek

-- 
                *8^)

Email: [email protected]

Get CVS support at <http://ximbiot.com>!
-- 
125. The truth is out there? Anyone know the URL?