Re: "weird" naming convention for ocamlbuild executables
On Tue, 2007-06-19 at 10:15 +0100, Stefano Zacchiroli wrote:
> > You can't do that, it will break things. In Felix at the moment,
>
> This is a way too bold statement. We *can* do that, we are software
> distributors
Well yes, you CAN do it .. I didn't mean you physically can't do it,
I meant I don't think it is advisable to change the names of executables
from what Inria names them.
> and we do this kind of stuff if we think this way the
> software will better suite our distribution. If it breaks (guess what)
> Felix we will patch it so that it won't break.
You're assuming ONLY Debian packages depend on Ocaml.
This is not the case. People actually write their own non-packaged
code, and you can't (sic) patch that.
> The majority of applications won't break. And if INRIA is going to adopt
> the .byte/.native convention anyway, the few patches we will need to
> introduce for that will vanish in next releases.
If Inria is going to do it anyway, my argument evaporates.
> > The difference between ocamlopt.opt and ocamlopt for building debian
> > packages is irrelevant anyhow. Who cares if the Debian autobuilder
> > takes a few extra seconds to build something?
>
> I don't know, let me think: what about the buildd administrators when
> the daemons are overloaded just before a release?
A few extra hours build time for the Debian
build system hardly compares with the many end users developing
code for Ocaml every day.
Sorry .. I don't care about the buildd admins. I'm more concerned
with the quality of my product. That's why Felix takes maybe 60
seconds to compile .. and 20 minutes to test .. and the tests
are all part of the build. Switching to ocamlopt instead of
ocamlopt.opt isn't going to make a dent on the buildd time:
the tests are run anyhow.
If I had my way, I make unit testing all Debian packages
mandatory. No test, no package. And that would certainly
greatly increase build time. I don't care .. it would
also greatly increase the quality of the product
(IMHO of course -- it may not be so, but at this stage
I believe that).
--
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net
Reply to: