Hi everybody,
> 1. Linux, by its design, is a multi-task system, and there's nothing in
> the Policy that disallows spawning background tasks from maintainer
> scripts, so I don't understand why this bug was marked serious.
Actually, I have to agree on that one. (Although I'm glad it got this
much attention in the first place :p )
> 2. `true' is a sensible default for the question for interactive use
> (when user actually sees the question). For non-interactive use, `false'
> would be better. I'm thinking about distinguishing between interactive
> and non-interactive usage by using debconf's seen flag - i.e. skip the
> indexing if user hasn't seen the question.
>
> 3. There's another task spawned (unconditionally) by dwww's postinst:
>
> if ( -x "/etc/cron.daily/dwww" )
> {
> print STDERR "\nBuilding dwww pages in the background...\n";
> system("setsid /etc/cron.daily/dwww &");
> }
> This should take a far less time to complete, but still there's a
> similar race condition to the one reported here. So probably this also
> should be run in foreground in non-interactive installation
Definite ACK to the latter one.
If you could achieve such a distinction, this would already help very
much. OTOH still this wouldn't constitute a general basis.
@Daniel Bauman:
Why do you think performing postinst tasks in background is "wrong on so
many levels"? As long as the setup is performed interactively and the
user gets notified accordingly, there shouldn't be an issue.
Non-interactively on he other hand, quite a bit of additional
precautions would have to be taken, no doubt. So: what's the correct
procedure here? Should this really be taken up to extending the Debian
policy?
Cheers
Daniel
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
