Re: [Boost-testing] Bug in regression.py?

From: Markus Schöpflin (markus.schoepflin_at_[hidden])
Date: 2005-11-30 03:43:12

Markus Schöpflin wrote:
> Rene Rivera wrote:
> 
>>Markus Schöpflin wrote:
>>
>>
>>>Hello all,
>>>
>>>the function update_source() in regression.py currently looks like this:
>>>
>>>def update_source( user, tag, proxy, args, **unused ):
>>>    if user is not None or os.path.exists( os.path.join( boost_root, 'CVS' 
>>>) ):
>>>        open( timestamp_path, 'w' ).close()
>>>        log( 'Updating sources from CVS (%s)...' % timestamp() )
>>>        retry(
>>>              cvs_update
>>>            , ( user, tag, args )
>>>            )
>>>    else:
>>>        get_source( user, tag, proxy, args )
>>>
>>>Obviously the first line is supposed to check whether a CVS update can be 
>>>done. For this to work correctly, I think the 'or' must be changed to an 
>>>'and', right?
>>
>>
>>No. The check should be equivalent to: if there's a CVS user passed in 
>>*or* if there's a CVS user already encoded in boost-root/CVS/Root file; 
>>then one can do a CVS update. Perhaps it should be change to 
>>specifically check for the CVS/Root file instead of just the CVS subdir.
> 
> 
> But you can't run CVS update when no CVS folder is present, can you?

Thinking some more about it, the code should do the following IMHO:

If there is a CVS/Root file present, then call cvs_update(), no matter if a 
user is given or not. Otherwise call get_source() which itself already 
checks if it can use CVS to get the sources.

IOW, the 'user is not None or' should be removed, and it should check for 
<boot_root>/CVS/Status instead of <boost_root>/CVS. D'accord?

Markus