Autoconf 2.60 is not released yet, though we are very close. Thus we cannot require it.

Let me explain the original problem:
With Autoconf 2.60 generated configure, @datadir@ will be replaced by ${datarootdir}
This means that the `datarootdir' variable will have to be in *.pc.in templates, for the same reason for which current *.pc.in contains `prefix', `exec_prefix', etc.

In Autoconf 2.59, `datarootdir' does not exist, so the string @datarootdir@ remains intact.

Even though it would be easy to make provisions for removing the line, or for setting detarootdir to ``this variable is not used by this version'', I don't think it is worth it, unless a real problem appears.

My current understanding is that this is only a cosmetic issue, which will fade out soon, as Autoconf 2.60 will take the world.  ;-)

I understand what's going on, but not why it's implemented in bonobo *now*. That is, when 2.60 is used, will *.pc really "have to" have a datarootdir variable?

I don't mind having it left as @datarootdir@ if it doesn't break anything presently. It was something flagged while validating my *.pc collection...had over a dozen unsubstituted @*@ strings, this (same as in Bug #342327) is the only case where it wasn't due to a clear typo in .pc.in or brokenness in configure.in.

(In reply to comment #2)
> I understand what's going on, but not why it's implemented in bonobo *now*.
> That is, when 2.60 is used, will *.pc really "have to" have a datarootdir
> variable?

Well, the answer has evolved in time, so I'll give you two answers instead of one:

As of 2005-08-16, when I committed the patch:
If the (then) current CVS version of Autoconf was used, the resulting *.pc contained line
  idldir=${datarootdir}/idl/bonobo-2.0
(supposing that the --datadir option was not used).
Had the *.pc line lacked the datarootdir variable, this above line would imply creating the /idl directory under the root.[*]

So back then, the answer was clearly: "yes, the *.pc absolutely have to contain the datarootdir variable."  And I was not aware of any plans to change the situation in further development.

The situation, as of now, is different:
If the template file contains the string "@datadir@", but not the string "datarootdir", configure issues a warning and changes the value of datadir do that it doesn't contain the reference to datarootdir.

This hack means that the datarootdir is not so strictly required, yet I think that the cosmetic problem of a "WARNING" from Autoconf-2.60-generated configure is bigger than the cosmetic problem of having "@datarootdir@" in the *.pc file. Consequently, I think that this change should be done rather sooner than later. And I beleive that Autoconf 2.60 will be out before summer 2006.

> I don't mind having it left as @datarootdir@ if it doesn't break anything
> presently.

I do not think it causes any problems.  I have made this change to several Gnome projects back then, and this is the first complaint I have encountered.

> It was something flagged while validating my *.pc collection...had
> over a dozen unsubstituted @*@ strings, this (same as in Bug #342327) is the
> only case where it wasn't due to a clear typo in .pc.in or brokenness in
> configure.in.

This is a very useful work, thank you very much for that. I apologize for the complications caused by my changes.

I have two ideas for your validator, probably obvious:

1) if the *.pc file contains ${foo}, but does not define the variable `foo', this is a bug. (Does pkg-config complain in such a situation?  I think it should.)

2) if the definition of `foo' comes after the ${foo} occurence, how does pkg-config cope with it?  (Or other programs which parse *.pc.)