Re: calc.at workaround for current test failures

[Kaveh R. Ghazi]

 > From: Paul Eggert <address@hidden>
 > 
 > Frank Heckenbach <address@hidden> writes:
 > 
 > > Anyway, is there a problem with making that change?
 > 
 > I'd rather not make a change because GCC has an inaccurate warning.
 > We should fix GCC's warnings instead.
 > 
 > I'll CC: this messaage to Kaveh R. Ghazi, since he's the one who made
 > that change to GCC (logged in
 > <http://gcc.gnu.org/ml/gcc-patches/2000-07/msg00267.html>).
 > 
 > Kaveh, here's some context:
 > 
 > http://mail.gnu.org/archive/html/bison-patches/2003-07/msg00013.html
 > 
 > I am worried that GCC is diagnosing more than it should.  Did you
 > check whether traditional C rejected "struct s a = b;"?  I know that
 > it rejected "struct s a = { x, y};" but I don't recall that it
 > rejected "struct s a = b;".  Nobody in the Bison group has an old PCC
 > compiler to test with, so we can't verify this ourselves.

Sorry, I no longer have access to any K&R cc systems.

When I did have a K&R cc, it was primarily SunOS4 cc.  I also saw
reports from hpux cc users and used some of those to determine what
warnings to add.  I seem to recall that one or both of these compilers
rejected both "struct s a = { x, y};" and the "struct s a = b;" form.

Rather than rely purely on my memory, I scanned through the GCC
ChangeLogs prior to July 2000 when I installed the new -Wtraditional
warning.  I looked for fixes of automatic aggregate initialization
which should have been inspired by actual old compilers rather than my
warning patch.

I found this from November 1999:
http://gcc.gnu.org/ml/gcc-patches/1999-11n/msg00777.html

It's actually "struct s a = func();", nevertheless it lends support to
the fact that the GCC warning is correct.

                --Kaveh
--
Kaveh R. Ghazi                  address@hidden