[companion-discuss] Re: open code for fundamental libraries
Dennis Clarke
dclarke at blastwave.org
Tue May 2 14:51:09 PDT 2006
>
> lianep at eng.sun.com writes:
>>
>> Eric Boutilier writes:
>> > My initial thought is that not all the items seem to fall in the
>> > cateogory of common source repository requirements; some seem to be
>> > build-system requirements. (On the other hand, the problem could be that
>> > my notion of what a source repository is is way too narrow...)
>>
>> That's a fair comment about the terminology. But...
>>
>> A source repository with absolutely no build infrastructure is probably
>> best implemented as a web page with links to the main source download
>> sites for lots of OSS components. That way, you know the components are
>> always up to date. :)
I was in the process of putting together some build infrastructure.
The usual pieces, build systems and central storage.
>> If the primary problem worth tackling as a community is to make
>> sure that the main trunks of the OSS components can build on
>> OpenSolaris-based distributions, I think we'd probably be better off
>> eschewing the overhead of a source repository and just creating a wiki
>> which had a list of OSS software and interesting versions with
>> instructions
>> on how to compile. (Bart suggested this -- it wasn't my idea.) If we
>> can't reach a consensus on a useful build environment, pursuing such a
>> wiki would be a great idea.
>
> (I should also note that I didn't miss Dennis' inventory of libraries
> in this thread, which is a great start down a similar path. I wasn't
> trying to claim Bart was the only or even the first one to have this
> idea.)
I think we are all having the SAME ideas at the same time. :-)
>> It just seems like a common build environment would even more helpful as
>> it has the benefit of turning those manual instructions into a set of
>> makefiles so the instructions can be reduced to "make install" (or a
>> more appropriate target) for each component rather than differing for
>> many of them.
Perhaps we can arrive at some ideas and solutions through the process of
experimentation. What if we were able to simple create a snv_38 based build
server set and then use them to experiment our way towards a solution? I
guess I am just a kinda hands on guy.
oh .. also .. I am reading over your email with your list of criteria and
spec suggestions and trying to type up some intelligent responses.
--
Dennis Clarke
More information about the companion-discuss
mailing list