You are viewing a plain text version of this content. The canonical link for it is here.
Posted to general@gump.apache.org by Nicola Ken Barozzi <ni...@apache.org> on 2003/04/24 11:58:20 UTC
Question on profiles
We have now basically one profile, ./profiles/gump.xml.
It would seem logical to make for example a "krysalis" profile for those
that only want to build krysalis projects with Gump.
http://jakarta.apache.org/gump/overview.html :
'A profile is a collection of projects and repositories that are
logically to be considered a unit. A profile can provide a consistent
set of installable dependencies, or tagged versions of cvs modules to be
used.'
Then it seems logical to remove the krysalis definitions from gump.xml
and simply import two profiles in the workspace.
But... if I do it, I will find myself also with the installed jars being
present in the merge...
I read:
'The way extending or overriding works is quite simple. If two like kind
elements (e.g., project) are located with the same value for the "name"
attribute, they are combined. Combining of attributes results in like
named attributes overriding previous values. Combining of elements
results in concatenation. These rules allow for the easy overriding of a
target and the adding of a dependency.
An area for future exploration is "anti-elements", which cancel out the
effect of other elements'
Sam, I am confused WRT the actual use of a profile, and I guess that I'm
overloading it's significance. Nevertheless, you suggested somewhat to
make a profile for avalon when we were talking about building
avalon-only with Gump.
Could you please give me insight?
TIA
--
Nicola Ken Barozzi nicolaken@apache.org
- verba volant, scripta manent -
(discussions get forgotten, just code remains)
---------------------------------------------------------------------
Re: Question on profiles
Posted by Sam Ruby <ru...@apache.org>.
Nicola Ken Barozzi wrote:
>
> What makes me think, though, is that basically the "installed packages"
> part is the most delicate part in a Gump install.
+1
-Sam Ruby
Re: Question on profiles
Posted by Nicola Ken Barozzi <ni...@apache.org>.
Sam Ruby wrote, On 24/04/2003 15.06:
> Nicola Ken Barozzi wrote:
>
>>
>> Sam, I am confused WRT the actual use of a profile, and I guess that
>> I'm overloading it's significance. Nevertheless, you suggested
>> somewhat to make a profile for avalon when we were talking about
>> building avalon-only with Gump.
>
> I originally was thinking in terms of there only being one profile that
> a given workspace would use. An Avalon or Krysalis profile would simply
> contain the versions of the projects that made sense to those developers.
Hmmm, but an Avalon developer is not /only/ an Avalon developer, that's
the problem. This phrase instead makes sense if we talk about an
Avalon-Gump-Server-Install, that does automatic Gump only for those
projects. Hmmm...
> The idea of multiple profiles is an intriguing one. But it does bring
> up the question in general about how to resolve the conflicts. I'm not
> sure I have any good answers here.
What I tend to think is that installed jars should not be specified in
profiles. In this way we could simply have:
profile/krysalis.xml
profile/apache.xml (servers)
profile/apache-avalon.xml
profile/apache-jakarta.xml
profile/apache-cocoon.xml
...
etc
In this way users can mix and match the compile "units" they prefer.
This is now reflected in the informally <!-- commented --> sections in
the gump.xml profile.
Here conflicts must not arise, because each unit has to be completely
indipendent. If I import 10 profiles in my workspace or just one, the
thing should not change on how it's treated, just as I would "xmlcat"
all these profiles before inclusion.
What we could then also do is to add a
installed-packages.xml
profile that only exports installed packages.
In this way a I could just add that to my profiles.
What makes me think, though, is that basically the "installed packages"
part is the most delicate part in a Gump install.
For instance, if I add certain "split" profiles to my system, I would
assume that I want those to be gotten from CVS and the rest done as
installed packages.
Some users would be happy with an automatic
installed-packages.xml
profile creation, others with a manual one.
In any case, then a nice feature is to have the following behaviour:
1 - packages.py is called
2 - it checks all the jars that are needed to complete a full run
3 - it outputs a installed-packages.xml file in a certain
location (where?)
Then the user can look at it, tweak it, check it, etc. The workspace ha
sto include it too of course as an import.
Then we have
4 - ruper.py updates the jars from what specified in the workspace
(not necessarily jars are specified only in installed-packages.xml)
5 - Gump can run
It still requires users to manually do steps, but it's all guided.
Does this make sense?
--
Nicola Ken Barozzi nicolaken@apache.org
- verba volant, scripta manent -
(discussions get forgotten, just code remains)
---------------------------------------------------------------------
Re: Question on profiles
Posted by Sam Ruby <ru...@apache.org>.
Nicola Ken Barozzi wrote:
>
> Sam, I am confused WRT the actual use of a profile, and I guess that I'm
> overloading it's significance. Nevertheless, you suggested somewhat to
> make a profile for avalon when we were talking about building
> avalon-only with Gump.
I originally was thinking in terms of there only being one profile that
a given workspace would use. An Avalon or Krysalis profile would simply
contain the versions of the projects that made sense to those developers.
The idea of multiple profiles is an intriguing one. But it does bring
up the question in general about how to resolve the conflicts. I'm not
sure I have any good answers here.
> Could you please give me insight?
>
> TIA
>
>