This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

gengtype vs ObjC++


While everybody else gets to work on compiler speedup :-), I'm stuck
with the somewhat mundane task of getting our ObjC++ implementation
to work with gengtype machinery.  Conceptually, it's simple; ObjC++
just needs to be able to trace all the roots in Objective-C, plus
all the roots in C++, effectively a union of gtype-objc.h and
gtype-cp.h.

However, gengtype seems to be very big on the idea of having
gtype-<lang>.h synthesized from a <lang>/foo.h file.  But if I make
a synthetic objcp/all.h made by catting cp/decl.h etc plus
objc/objc-act.h, I get a bunch of duplicate structure errors, and
even if I dike out the error reports in gengtype, the result is still
missing all the root tables for C++, so I'm still being outsmarted
somehow.

Is this a plausible approach, or should I look for something else?
Gluing gtype-objc.h and gtype-cp.h together seems hard because
of the way the root tables are built up.  Could it work to have
the ObjC roots always in cc1plus, but be empty/undefined?

Stan



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]