Public SourceMage Discussion List

[Seth Alan Woolley <seth AT positivism.org>]

On Sat, Sep 03, 2005 at 05:44:16PM -0500, Jeremy Blosser (emrys) wrote:
> On Sep 03, Flavien Bridault [f.bridault AT fra.net] wrote:
> > Le vendredi 02 septembre 2005 ? 20:56 -0700, Seth Alan Woolley a ?crit :
> > > In the future, we can pick one place to put notices that stuff is risky 
> > > instead of having people check irc, ml, etc.  I suggest that not be the 
> > > HISTORY file nor the Changelog, otherwise a false alarm might work its 
> > > way up the grimoire to stable, and I'd rather not have to filter every 
> > > changelog-like file for stuff that's not relevant anymore.
> > > 
> > > I think a note to the mailing list would be best.  It's the best place 
> > > currently, and Sergey made the right decision here.
> > > 
> > 
> > Just a question, what will you do for the integration into stable-rc and
> > then in stable ? You will integrate it as it is now ? Without any notice
> > in the spell ? So, if you include one, or any mechanism, why not do it
> > at the beginning in devel ? Is that really too painful ? Trying to keep
> > test grimoire as much stable as possible is anyway the better thing to
> > have a good stable grimoire then.
> 
> I agree that any notices that are going to go out should start as early as
> possible, if for no other reason than to let us "test" the warnings
> themselves for effectiveness.  However, if they only happen on the mailing
> lists, this can become spammy.
> 
> I don't think they should just happen on the ML.  Too few users read it,
> and that problem will only increase as time goes on and our user base
> increases.  Even those that do read it may go months without time to keep
> up, but that doesn't mean they stop updating their systems.  The only thing
> we know they have is the spell, and a warning if it's necessary should go
> there.  Of course, ideally we can figure out ways to not have to warn them,
> and some of those have been suggested.
> 
> > However I'm guilty, I must admit, because I didn't know that openssl
> > 0.9.7d->0.9.8 update will imply an API change. When a gcc or glibc
> > update comes, its true that I'm much more careful, regardless to the
> > eventual posts in the ML. The next time I will know that. ;-)
> 
> gcc, glibc, openssl, and zlib are the big ones in my experience.

After talking with Flavien in IRC, I have decided that a HISTORY entry 
along the lines of the following would sufficed for both of us:

"Please note that 0.9.7x and 0.9.8x have a different ABI and that a 
recompile of all openssl dynamically-linked applications will need a 
recompile before they will work properly.  A cleanse --fix should find 
these instances and automatically repair them for you."

As long as we don't mention the specific process of integration, we 
should have no problems with putting notices in HISTORY entries like 
this as far as not confusing users.

Perhaps we assumed people already knew this -- maybe we shouldn't in the 
future until we get ABI-aware triggers in place to handle it without a 
cleanse --fix.

Seth

-- 
Seth Alan Woolley [seth at positivism.org], SPAM/UCE is unauthorized
Quality Assurance Team Leader & Security Team: Source Mage GNU/linux
Linux so advanced, it may as well be magic http://www.sourcemage.org
Secretary Pacific Green Party of Oregon http://www.pacificgreens.org
Key id FDCEE733 = 5302 B414 64C4 6112 3454  E082 99F0 69DC FDCE E733

Attachment: pgp1_YHKkk3hn.pgp
Description: PGP signature