gump-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefano Mazzocchi <>
Subject Re: [Cactus] commons-codec added to dependency jars?
Date Sat, 13 Mar 2004 05:30:04 GMT
Vincent Massol wrote:

>>-----Original Message-----
>>From: Stefano Mazzocchi []
>>Sent: 12 March 2004 21:09
>>To: Gump code and data
>>Subject: Re: [Cactus] commons-codec added to dependency jars?
>>Vincent Massol wrote:
>>>The way I code is as follows:
>>>1/ make a change on my local machine
>>>2/ verify it works by running some kind of tests
>>>If I don't do 2/ then there is a high chance it will break something
> and
>>>I'll only know the following day.
>>>To do 2, there are 2 solutions I can see:
>>>1/ install gumpy locally but that's not very easy to do
>>>2/ have a kind of "self-service gumpy" that you can start on demand
>>>What do you think?
>>I never though that people would have a problem modifying the metadata
>>files for fear of breaking gump.
>>But, I wonder, if your build is already broken, why are you afraid of
>>breaking it?
> In this specific case, it's not broken. 

oh, ok.

> It's a refactoring (replacing
> the dependency on commons-code with some inherit on httpclient). The
> problem is that I'm not a gumpy expert and I don't know if this change
> is the correct one. Stefan, on the other hand, has had much more
> experience and I'm sure he's more confident in the changes he's making
> (even though, I'm sure he tries them on his machine from time to time
> before or after committing the changes) :-)
> More generally, if the build is broken and I make a change to fix it,
> I'd like to be sure it fixes the build. Otherwise I'll have to wait till
> the following morning, try something else, wait again, try something
> else. I remember doing this in the and it took me about 1 week or
> more to get Gump working for Cactus (the Cactus build does quite a lot).

man, it took me two months to have a single clean gump run for cocoon, 
then I let go and it fell apart, so I do feel your pain.

I agree that a sandbox to play would be nice, like having a workflow like:

  1) you log in
  2) click "edit the project descriptor"
  3) a textarea contains the descriptor
  4) you change the descriptor
  5) save it
  6) run the descriptor
  7) get the output
  8) go back to 2) until you are satisfied
  9) commit to cvs

Two problems in this scenario:

  1) infrastructure is not going to allow gump to commit to CVS directly
  2) since there is no LDAP or equivalent, it's tricky to have a 
committer-wide access control in synch [even if infrastructure would 
like to have a single sign-on system in place]


View raw message