[CVS gcc toplevel patch] v2 + v3, --enable-libstdcxx-v3

[Martin v. Loewis]

> Will this enable -fsquangle also? 

Enabling -fsquangle as a stop-gap measure might be reasonable. In the
long run, it turns out that -fsquangle is dead code.

> I saw some contrived examples in which code generated was almost an
> order of magnitude more without -fsquangle. I know this will break
> binary compatibility but I wonder who uses libstdc++ without knowing
> all that. Maybe even -fnew-abi and all those other options which get
> GCC on par generating the best code for std C++? Or is this going to
> be too risky to turn all that on?

I don't know how risky -fnew-abi is; I haven't tried in a
while. However, it will change on a daily basis, so yesterday's
libstdc++ might not work with today's compiler, then (even without any
change in libstdc++). Also, the mangling mechanism of -fnew-abi will
be significantly different from -fsquangling, or anything else that
gcc had before (well, 'i' still means 'int' :)

Also, there is the libio problem that the new ABI is incompatible with
Linux glibc/libio. I don't know whether this has been resolved, yet.

Regards,
Martin