This is the mail archive of the 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]

Re: g++ and gcc behave differently

>> Yes.  Feel free to fix it.  Though I don't know which one we want
>> to be the default.  Actually one could argue that the way g77 -v
>> works ought to be the default
>Ouch, oh man, is that one nice.  :-) It wouldn't be bad if we all did
>it that way.

I think before we try to copy g77's approach around to other
drivers (after 1.1, obviously ;-), we should discuss how we
really *want* these things to work, vis-a-vis their design.

E.g. now that I've seen `gcc --help' vs. `gcc --verbose --help',
I think maybe `gcc --version' should perhaps produce what
`gcc -v' now does, and `gcc --verbose --version' what `g77 -v'
now does, except perhaps it'd invoke *all* the drivers or only the
cc1 instead of the f771.

Seems to me `gcc -v', being shorthand for `gcc --verbose', should
just produce a version number and then "No input files" or whatever
`gcc' by itself would produce.

That's just one suggestion, though, and unfortunately it is
incompatible with past behavior and expectation, although I think
it's more compatible with GNU Coding Standards (which I haven't
read for some time).

Once we agree on a design, we can then worry about how to implement
it portably.

At that point, perhaps some of g77spec.c's internal differences
vis-a-vis g++spec.c's can be made use of one way or the other.
E.g. some of g77spec.c's abstractions might be worth hoisting into
gcc.c itself, making the *spec.c files smaller or, at least,
easier to maintain.

        tq vm, (burley)

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