bootstrap failure, gcc HEAD, --disable-shared, gnatlib-plain

[Arnaud Charlet]

> Arnaud, I'm reverting much of your recent change to libada; it's causing 
> this.  In particular, the collapsing of multiple targets into one with 

Right, although it was trivially fixable.
Also, having duplicated ajd almost identical code is also a good source
of bugs, so I disagree with the above.

> $@ was done wrong.  Apart from the error above, there's also the sudden 
> removal of GNATLIBLDFLAGS from the flags passed to gnatlib-shared. 

Also easily fixable.

> Futhermore, I don't actually think they should be collapsed at the 
> moment; it invites further bugs.  :-P

I disagree, it reduces the potential for bugs and discrepencies.

> Apart from that, there are the stylistic issues; overuse of GNU make 
> syntax is significantly less maintainable than using 'configure' 
> properly.  ;-)

I also disagree, I find the use of GNU Make very handful.
I just believe you (and other GCC folks) are not yet ready to use these
features, that's all :-)

Having used GNU Make for GNAT for years now (compare to a very recent
switch in that direction for the rest of GCC), I find it very valuable,
and find overuse of configure not necessarily preferrable, in particular
since configure is by itself a big hack which is painful to maintain :-)

Anyway, I don't have any particular problem with your proposed patch.
What I really didn't like in your previous change was the overuse of 'sh'
embedded in Makefile, which *is* very error-prone and painful to maintain.

Arno