Justin Erenkrantz wrote:
> --On Friday, January 21, 2005 9:49 PM +0000 Max Bowsher <maxb@ukf.net>
> wrote:
>
>> If I run "make install-some-additional-feature", I expect
>> "some-additional-feature" to be installed. I don't expect
>> "some-additional-feature", and some _subset_ of the base package to be
>> installed.
>
> IMHO, there's two legitimate viewpoints on this:
>
> 1) install dependencies
> 2) do not install dependencies
>
> We originally did #2, but had a significant number of bug reports asking
> us
> to do #1. So, we modified the make system to install those dependencies
> as
> well - because people claimed that they expected 'some-additional-feature'
> to be all that was necessary to install that additional feature. (They
> didn't want to run 'make install.')
>
> *shrug*
>
> I think we're in a no-win situation here. So, I'm against changing it one
> more time: whatever we do is going to get complaints. Best to leave it as
> it is. -- justin
I'm for changing it again, because:
1) I think we should be phasing out the direct use of install-foo by users,
in favour of configure --enable-foo flags, because that is what people tend
to expect in this autoconf-dominated world.
2) The previous change to the current situation removed functionality useful
for packagers, or people locally installing in interesting configurations,
and left no easy workaround. Whereas people wanting dependencies ought to
first understand that their binding of choice uses the subversion libraries
which may be used by other software too, so should be implicitly replaced,
and also, they have the trivial workaround available of "make install
install-foo".
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Jan 22 00:20:52 2005