On 11/3/06, Philip Ganchev <[EMAIL PROTECTED]> wrote:
> On 11/3/06, Axel Liljencrantz <[EMAIL PROTECTED]> wrote:
> [...]
> > I think that forgetting to do a make uninstall, or even losing the
> > source so you can't do one is pretty common. I've done it myself a few
> > times. :-/
> >
> > I've added a note to the INSTALL file about uninstalling older fish
> > versions, hopefully that will help. I've also added a script that
> > checks if the seq command is a fish script, and if so, replace it.
> [...]
>
> Could the uninstall be semi-automated? Something like:
>
>
> It seems that your computer already contains files from a fish
> installation, apparently version 1.21.0. They must be uninstalled
> before doing another install. Inside the directory containing the
> source code of fish 1.21.0, type
>
> sudo make uninstall
>
> You can get the source code of fish 1.21.0 from http://fishshell.org,
> if you do not already have it.
Unfortunatly, that doesn't work. If you redownload the source, you
will rerun the configure script, and it will detect that you already
have a seq version, and it won't try to install a new one, meaning
'make uninstall' won't remove it either.
The only two options are either to uninstall using the _original_ fish
configuration that installed seq, or using the script that I've added
to configure that tries to detect if the installed seq version really
comes from fish.
The question then becomes what should the script do if it detects an
old seq version. It can either fail with an error message telling the
user what went wrong, or it can decide to overwrite seq. I personally
think that fish should not fail and tell the user to manually remove
seq, since fish does not do that if you e.g. already have a fish
version installed. In that case, fish will happily overwrite the old
fish version, so why should it be different with seq? It would be
completely inconsistent. If you run 'make install', you should be
prepared that the old version of your program will go the way of the
dodo.
>
> Ideally, there would be a script that tries to configure, make and
> make install automatically and intelligently -- say if it is run
> without root privilige, ``configure --prefix=$HOME''. It breaks Unix
> convention, but that convention is not very friendly or interactive.
> I'm not sure what that script should be called, because the best name
> is already taken by a script ("install-sh") that does something else.
The three commands './configure' 'make' and 'make install' can be
replaced by the two commands './configure' and 'make install' if you
feel lazy. You really should check that the configuration step went
well before building, though, so I don't like the idea of reducing
this to only one step. Also, I think that every user that wants
point-and-click class ease of use while installing will be better of
using a graphical package manager. The source build process should be
as easy and streamlined as possible, but let's not forget who the
target audience is.
--
Axel
-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Fish-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/fish-users
