------- Comment #3 from dave at hiauly1 dot hia dot nrc dot ca 2007-10-28 22:47 ------- Subject: Re: "error: array type has incomplete element type" in system header
> > In particular, the following DR was referenced as proof the above is > > "not valid C": > > http://www.open-std.org/jtc1/sc22/wg14/www/docs/dr_047.html > > > > It says there regarding the submitted examples that "First of > > all, no constraints are violated. Therefore, no diagnostics are required." > > A constraint was added in C99. Ok, but this wasn't with -std=c99. > > So, I'm questioning why a -pedantic warning was changed to a hard error > > in gcc 4.0.0. This seems to have been ignored in Joseph's response > > and the subsequent change. > > I referenced bug 19333 in my message you reference above; it's an ICE > showing the former undocumented extension did not work (the original bug > summary was "Compilation SEGFAULTs with -O1 -finline-functions on the > x86_64 architecture."). That in turn links to gcc-patches discussions > leading to it becoming a hard error. > > The system header issue could be fixed by fixincludes. Sure but it takes time. I imagine the testcase in bug 19333 would cause a problem on the PA. However, isn't the real problem the use in function `g'. When I look at the assembly code generated by gcc-3.4.6 using the testcase in this PR, there's no reference to syswait at all. This is what I would expect for an extern that isn't referenced. Dave -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33933
