bug#50469: [bison-3.8] bug or side effect to flex & automake

[Akim Demaille]

Hi!

> Le 9 sept. 2021 à 00:32, Paul Eggert <eggert@cs.ucla.edu> a écrit :
> 
> On 9/8/21 2:18 PM, Karl Berry wrote:
>> Just an idea that I don't expect you to adopt, but just to mention --
>> you could only institute the breaking change if POSIXLY_CORRECT.  That's
>> why POSIXLY_CORRECT exists. -k
> 
> I like this idea. It insulates us against POSIX decisions and/or indecisions 
> in this area.

Good with me.  I'll implement that, thanks Karl!


>> P.S. Is this related to Automake? If it is, as with Autoconf, there
>> needs to be new volunteer effort to do what's needed. I can't take it on
>> and unfortunately no one else has come forward.

IMHO the prototypes in Automake's test suite are wrong: I see no reason for
yyerror to take a char *, and we do pass constant strings to it, so I'm
quite surprised that there are no compiling errors about it.  We should
fix that.  I can do that.

On the Autoconf side, we dearly need some AC_CHECK_BISON.  It is really a
pity that everybody uses AC_CHECK_YACC, which seriously degrades the point
of using bison.

Cheers!