Torsten Landschoff wrote: > > The bug is in slapd for including this text in its debconf template: > > > > "The default is /var/backups/slapd-VERSION" > > How is that a bug? In fact this can be helpful in case you changed the > value and later wonder what is was originally. Apart from that it is an > example how to use the VERSION tag. > > > This comes under the heading of not referring to debconf UI in a > > template. Just as you don't know how debconf will choose to present a > > yes/no question and thus "say yes" constructions should be avoided, you > > don't know how or if a given debconf frontend handles default values[1]. > > So you suggest removing that string and leaving the user completely in > the dark what to enter which is utterly needed especially if the debconf > frontend ignores the default value.
There's nothing stopping you from including an example value in the
template. Don't present it as the default value.
> > Indeed a static template such as this one doesn't even know for sure
> > what the default value _is_; it could have been overriden.
>
> By whom? As I am the maintainer of slapd I expect nobody else to change
> that default.
Preseeding? Derived distributions?
> If you really think we should not mention the default value there that's
> okay but on the other side I request that the behaviour of the readline
> frontend is changed to at least present the user the current/default
> value if the libterm-readline-gnu-perl package is not installed. And
> make it very clear that hitting return means submitting an empty value.
It does. It prompts as follows:
Directory to dump databases: _
Barring some note that tells them there is a default, noone would expect
hitting enter to result in some default value here[1]. The convention is
that this means there is a default value:
Directory to dump databases [/var/backups/slapd-VERSION]: _
Or this:
Directory to dump databases: /var/backups/slapd-VERSION_
Debconf uses one or the other of these conventions when possible. Only
experienced users use these frontends, and experienced users are
expected to be aware of these conventions.
> Perhaps it's even better (for compatibility with the readline-installed
> case) to ask the user to explicitly input "" for the empty value.
And then users need to learn a complicated set of rules for the edge
cae where they want to enter a literal pair of double quotes. No thanks.
> > I'll reassign this back to slapd if it's agreeable.
>
> In case you do - what is the right action of slapd here? I'd rather
> avoid running each db_go in a loop which checks the input values.
Why? Validating user input is the correct thing to do in all cases
anyway.
--
see shy jo
[1] The only general exception to this rule I know of in all of unix/Debian
is bootloaders.
signature.asc
Description: Digital signature
