From: Phil Thornley on
On 25 May, 21:05, Yannick Duchêne (Hibou57) <yannick_duch...(a)yahoo.fr>
wrote:

[...]
> This latter one pass syntax check without any error message (just two  
> obvious warnings).
>
[...]

> Eh, I'm dreaming, I have the result I am reporting.
[...]

> I was doing a syntax check only (I feel it is important to state this, as  
> you come with an example using Derives clauses).

AAAAHHHHHHHH - I hadn't noticed that your OP said "syntax check" -
that is why you are not getting an error on the first example - it is
not a syntax error. (Presumeably because the preceeding declaration
*might* be an embedded package - which is allowed by the language
definition even if it's not implemented by the Examiner.)

(But it is a semantic error, as shown by my earlier response.)

Cheers,

Phil
From: Rod Chapman on
On May 26, 12:18 am, Phil Thornley <phil.jpthorn...(a)googlemail.com>
wrote:
> On 25 May, 21:05, Yannick Duchêne (Hibou57) <yannick_duch...(a)yahoo.fr>
> wrote:
> > I was doing a syntax check only (I feel it is important to state this, as  
> > you come with an example using Derives clauses).

Please don't use the "syntax check only" option - I wish we'd
never put it in there!

There are many many things which are allowed by the (context-free)
grammar used by the parser that are later rejected by the
semantic analyser - generic packages for one as you have
already discovered.
- Rod


From: Yannick Duchêne (Hibou57) on
Le Wed, 26 May 2010 09:38:49 +0200, Rod Chapman
<roderick.chapman(a)googlemail.com> a écrit:
> There are many many things which are allowed by the (context-free)
> grammar used by the parser that are later rejected by the
> semantic analyser - generic packages for one as you have
> already discovered.
Uh-oh, this seems to mean there is really no-no support for generics.
Unlike what I have believed I have read...


--
There is even better than a pragma Assert: a SPARK --# check.