Re: member with constructor not allowed in union

[Akim Demaille]

>>>>> "Anthony" == Anthony DeRobertis <address@hidden> writes:

Anthony> On Monday, March 18, 2002, at 06:28 AM, Hans Aberg wrote:

>> ... not compile the C output as C++.

Anthony> This used to work, in the same stable series of bison. It was
Anthony> never documented not to work, and perhaps is even documented
Anthony> to work.

That's correct.  `Accidently' is used to work (partially).  It has
never been implemented to this end.

Anthony> Changing that is fine and --- in view of the other C++
Anthony> problems using the C skeleton --- a good idea, but not in a
Anthony> stable point release.  It breaks a fair amount of
Anthony> software. It breaks stuff that, when reading the
Anthony> documentation, should work. Don't do that in a stable point
Anthony> release.

You seem to be drawing the wrong conclusion from the correct
observables: we did not mean to break it.  It just happened that some
people were using Bison in an unexpected way.  It is true,
nevertheless, that I was not aware of the number of people who play
this kind of trick to be able to use C++.  The fact is, I do myself
use (well, used to) a C++ parser with C output, but as I used %union,
everything was fine.  What we not expected, was that some people were
directly defining YYSTYPE.


Anthony> Do it in CVS. Do it for bison 1.5. It's a good idea, but for
Anthony> the right time.

Another good idea is C++ users using this CVS Bison.  As noted by
Hans, I want some tests first.  There are a couple of issues to fix
(for instance we still output #define, although enum are now easy to
output (Robert, would you handle this bit please?), and then, I plan
to give it a shot on some other projects than ours (e.g., GiNaC0, then
make a prerelease, and ask for opinions.


But people should understand that we _need_ input.  Do not wait for an
official release to realize it does not fulfill your needs!