On 2026/07/22 15:25, David Uhden Collado wrote:
> Stuart Henderson wrote:
> > we are pretty conservative in what is automatically run by pkg_add,
> > to my eye this is too much.
> >
> > I wouldn't object to installing a script and asking the user to run it,
> > though I think by default it should just print the operations needed so
> > the user can review (maybe with a flag to run them automatically after
> > checking).
> >
> > there should not be an @bin marker for scripts.
> >
> > the revised pledge still feels like shoehorning it into a program
> > which has not been designed to actually work with it. big clue
> > is having file access and network access in the same process.
> > (sure, there's worse in ports, and it's not a total blocker, but
> > it does make me wonder how well tested it's been, because clearly
> > there hasn't been too much careful thought about how to use this
> > prior to it being committed uptream..)
> >
> Hello Stuart,
>
> Earlier, you mentioned that users would not normally make these changes
> themselves, so the migration should be automated and made as robust as
> possible. That is what I have tried to implement.
I think you refer to this that I wrote earlier?
+-----
| > @newgroup _i2pd:838
| > -@newuser _i2pd:838:838::i2pd
account:${LOCALSTATEDIR}/lib/i2pd:/sbin/nologin
| > +@newuser _i2pd:838:838::i2pd account:${LOCALSTATEDIR}/i2pd:/sbin/nologin
|
| This will not get changed for users with existing installs
| so there will need to be some documentation about how to do that.
+-----
What I mean is: pkg_add will not automatically change this if the user
has previously installed an old version. (We generally consider this
too risky if the user changed something in a way that pkg_add / the
package doesn't expect).
We normally don't make changes like this in ports because of the extra
hassle for upgrading users. However if it is unavoidable then there must
be information for users to follow. I didn't say anything about
automating it (and actually making the changes automatically without
user review first seems risky in the same way it would be for pkg_add
to update /etc/passwd lines directly).
