C++ automatic template instantiation?

[Jason Merrill]

>>>>> Richard Kenner <kenner@vlsi1.ultra.nyu.edu> writes:

 >> I know too little about gcc's internals to really understand why
 >> Martin thinks -frepo is broken beyond repair...

 >     One problem is that there is a number of things that should work but
 >     don't, eg. producing libraries is difficult, and compiling with -S is
 >     not supported (as well as producing an executable directly from
 >     sources).

 >     The other (IMO more important) problem is that it will be always
 >     possible to trick the -frepo mechanism. For example, I believe
 >     recompilation currently does not take the GCC_EXEC_PREFIX variable
 >     into account (or any other environment variable). Also, never move any
 >     files around between compilation and linking, or recompilation won't
 >     find the original ones, anymore.

 > Perhaps I'm misremembering, but weren't these essentially the reasons
 > that RMS was so against putting in this code in the first place?

I think so.  But I felt then, and still do, that -frepo, while deeply
flawed, is better than nothing.

Jason