On Thu, Apr 14, 2005 at 09:50:42AM +0200, Torsten Landschoff wrote:
> > Apparently the upstream authors of OpenLDAP decided that
> >
> > limits user size=100
> >
> > is no longer accepted, meaning slapd fails to start after upgrade.
>
> It was never acceptable but the parser had a bug which caused undefined
> behaviour in that case. They forgot an else branch in a instruction of
> type
>
> if (strcmp(type, "...") == 0)
> ...
> else if (strcmp(...))
> ...
> else
> ...
>
> > It evades me why their parser in 2.1 didn't see this as a problem, yet the
>
> Because it was a bug?
>
> > new one suddenly breaks shit. I mean, if the old one ignored the line for
> > invalid syntax, why does he new one have to *die* when seeing it?!
>
> The same argument would apply to gcc which is much more strict in newer
> versions what syntax checking is concerned. I can't see how that is a
> bug.
Yes, except that gcc is usually used on demand and you don't get a downtime
when it changes its opinion on whether something is right or wrong.
And in other news, there really is little reason to treat "limits user" as a
broken phrase when "limits anonymous" is like that. I mean, if someone said
that it's for consistency, I'd like to know why the word anonymous isn't in
plural, too! :)
> > Please verify the integrity of ^limits lines in postinst upon upgrade,
> > like you do with other changes. TIA.
>
> I'll see if I find time to implement this. Currently I am only trying to
> fix things the old slapd.conf generated by debconf has as well and which
> really changed. I don't see why I should reimplement the syntax checker
> of the config parser in places where it works just fine now.
In the meantime, I've also seen slapadd die on things like:
telephoneNumber:: IA==
facsimileTelephoneNumber:: IA==
mobile:: IA==
Did the schema change to disallow a space, or did the checking become more
strict? No idea yet, but it also causes the upgrade to fail.
--
2. That which causes joy or happiness.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
