[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: critical problems in 2.3.5 prevent release

On Sat, Jun 09, 2001 at 01:35:16PM -0500, Stephen R Marenka wrote:
> > dpkg's handling of diverted conffiles i don't really know).  packages
> > in base should also not be asking questions in postinst, or blinding
> > running config programs *cough* quik *cough*.  they should at the very
> quik is a PITA.

no OpenFirmware is a pain in the ass.  but in any event i don't want a
quik bootblock installed whether the user wants/needs it or not, and
that is exactly whats happening now. 

> FWIW I'm not disagreeing with you, but is this a matter of policy or 
> personal taste? I don't recall reading about DEBIAN_FRONTEND or no
> conffiles in base packages in policy, but I could've missed it. Perhaps
> this is something that needs to be formally addressed somewhere or
> perhaps it is and I just need a pointer.

packages are not allowed to declare something a conffile and then go
screw with it in maintainer scripts or whatever, the reason is because
it causes dpkg to ask the user what to do with this conffile they
modified, when they the user NEVER modified it.  

since we are installing base onto a newly mkfsed partition there
should never be a modified conffile if dpkg says there is we have a
broken base package.  

as for questions in postinst, i think we need policy on that because
as it is you get VERY unexpected results.  (quik bootblock getting
installed with no config file as a part of `Install Base System').

> According to Richard Hirst, gcc-3.0 (at least for hppa) also has an
> issue.

why is gcc in the base system may i ask?  

> The way debootstrap is setup, that won't help, you have to look at the
> logs. debootstrap is running a repeat 20, more or less brute forcing the
> fixup of all the base packages (at least if understand correctly).

yes i have looked at debootstrap code, i still don't totally get it
all since you have to chase functions around in several files...  

but if these errors are to be expected and are the kind that work
there way out in the brute force method the user should not be getting
20 error dialogs.  if the last attempt fails again then the user needs
a messsage and preferably one with more info so we can actually try
and fix it.  

Ethan Benson

Attachment: pgpcgtaLPA9Vp.pgp
Description: PGP signature

Reply to: