[evla-sw-discuss] code organization

[Bryan Butler]

On 10/26/06 17:04, Pete Whiteis wrote:
> 
> 
>>>   Basically, I don't have any attachment to a particular version 
>>> management tool.   What would cause me pain is 1) the current 
>>> directory structure changed.  This could potentially force me to 
>>> change dozens of Makefiles, as well as utility scripts      
>>
>> well, the issue of whether it is a good idea to put hard paths into 
>> makefiles and scripts aside, how much work would it really be?  a 
>> day?  a week?  this is certainly an issue, i agree - we won't do this 
>> changeover for free - it will take some amount of work from everybody 
>> involved to get their code switched over (if we choose to do so).
>>
>>  
>>
> 
>    Of course its not a good idea to put hard paths into Makefiles. My 
> development platform/tools (Windows) doesnt have the concept of symbolic 
> references, nor do the tools play well with environment variables. We 
> were reluctantly forced into doing things this way.

understood.  though i thought that windows had that concept, i haven't developed 
any code on it, so trust you.

>    It would probably take a couple of days of dedicated time to flush 
> out all pathnames in both application and OS build files.

OK, this is a good data point.  we'll probably need to get estimates from 
everybody like this.

>>> 2) Losing tag information would be very bad.   I tag every release 
>>> I've placed into production since early 2004 and I've had to fall 
>>> back on at least 2 occasions.
>>>   
>>
>> whatever we did, we would always retain the old cvs repository.  if we 
>> switched to subversion, there are two ways to handle this:
>>  1 - try to get all of the old versions from cvs into subversion
>>  2 - start fresh on the subversion (only copy over the most recent
>>      versions), but keep the old cvs repository around
>>
>>  
>>
> #1 would be better, but initially painful.   #2 could be nasty if I had 
> to rollback near the cutover boundary. 

indeed.  we should get some input from folks who have done (or tried to) #1 
before.  i can't remember if the spitzer folks did this or not, but will ask joe 
chavez.  maybe somebody (brian?) can google it.  for #2, we'd have to try to 
have a relatively solid version for the initial copying.