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

Re: Blog post about package management (aimed at fedora)



On Tue, 2007-07-31 at 22:27 +0100, Paul Nasrat wrote:
> On Mon, 2007-07-30 at 08:12 -0800, Jeff Spaleta wrote:
> > On 7/29/07, Caolan McNamara <caolanm redhat com> wrote:
> > > FWIW OOo debuginfo is about 450 megs, which is rather sucky to force to
> > > download on a crash. And neither OOo or firefox are using bug-buddy. In
> > > the case of OOo at least the crash info can be re-annotated offline back
> > > to source lines, so we don't loose anything by having no debuginfo
> > > installed. Though we don't get local variables etc with offline
> > > annotation, but we don't have those with the current crash reporter
> > > anyway.
> > 
> > I think OOo illustrates that whatever toolized solution to help users
> > install debuginfo packages as needed, would need to be smart enough to
> > blacklist certain packages that are known not to benefit from the
> > debuginfo install.
> 
> I think the use case of users who can cope with submitting a bug with
> debuginfo are negligible compared to some other use cases. Plus most
> stuff will need to pull in library debuginfo deps, even if it's just
> glibc*debuginfo* to make sense of things.

More to the point, asking the user to install debuginfo _ever_ is a
failure.  You have the NEVRs of all packages involved in the crash.
Feed them to the bug machine, let the distro reconstruct the backtrace.
That way not only does the user not have to download $huge, but they
don't have to go hunt around in koji for the debuginfo that corresponds
to whatever outdated version of a given package they still have
installed.

- ajax


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