Re: [RFC] improve autotest syntax checks

[Ralf Wildenhues]

On Fri, Mar 07, 2008 at 10:41:12PM -0700, Eric Blake wrote:
> According to Ralf Wildenhues on 3/7/2008 11:30 AM:
>
> | m4_init([cleanup]) and AS_INIT([cleanup]) must belong to the least
> | intuitive interfaces that I've seen so far.  Why something named init
> | receives a cleanup argument rather than the name of or the thing that
> | it initializes, will not be understandable to users, and any apparent
> | analogy to AC_INIT or AT_INIT is just borked.
> |
> | I don't have a good idea how to improve this, but I must confess that I
> | don't like it much.  Maybe have a macro m4_init_wrap for this or so?
>
> In other words, guarantee FIFO behavior for m4_wrap (even with earlier M4
> versions) and add m4_init_wrap with LIFO behavior?  Sounds better than
> requiring registering the wrapup hook during initialization.  I'll work
> on that idea, and resubmit the series for review.

That sounds like an improvement

> |> +AT_CHECK_AT_SYNTAX([Multiple AT@&address@hidden,
> |> +[[AT_INIT([suite, take one])
> |
> | You need to double-quote comma inside AT_INIT's argument.
>
> Actually, I added a use of m4_expand inside AT_INIT so that I didn't have
> to double-quote the comma.  But as that was in a syntax failure test, I
> guess I should use a comma in a different test expected to compile cleanly
> if I intend to make that behavior the rule.

Hmm.  I remember to have seen a warning from AT_INIT about an extra argument.

Cheers,
Ralf