Le Lundi, 9 fév 2004, à 11:16 Europe/Zurich, Eric Knauel a écrit :
BTW, I'm not sure what --dry-run original intention was, seems like a
good option for debugging the installation procedure, so this little
patch to install-lib.scm simply prints load.scm to stdout if --dry-run
is given:
Nice! The only thing which slightly disturbs me is that it is
associated with the --dry-run option. For me this option is meant for
the end user, who wants to look at what would happen at package
installation time. It basically corresponds to "make -n install", which
is something I often use (--dry-run is the long name of -n in GNU make).
My impression is that the end user does not want to see the contents of
the load script when using --dry-run. It's more, as you say, something
that the package developer wants to use to debug his package
definition. So maybe it would be better to add yet another option, say
--debug, which would do several things similar to this in order to help
debug the package definition.
[In fact, in the long run I'd like to provide some additional script to
test package definitions. For example, I think it's very easy to make a
mistake when using the get-directory function and request the
installation-time directory when what's needed is the running-time
directory, or the other way around. That's hard to find out since
during most installations these are the same. And there are many other
things that one can imagine testing, like "is the load.scm file
created?", "are there really no files created when --dry-run is used?",
etc.]
So, if you don't mind, I'll associate your patch with a new --debug
option and not change the current behavior of --dry-run.
As far as your patch for CVS files is concerned, I also think that it's
something useful during debugging, as such files should not appear in
the final archive anyway. I'd also enable it only with the --debug
option.
Thanks,
Michel.
|