I'm debating whether llvm should support the "a.out" convention for the
default name of an executable file if no -o option is given to it. I'm
currently of the mind that this naming convention is completely outdated
and not worth propagating in llvmc. llvmc would require a -o option or
you get an error.
I'm debating whether llvm should support the "a.out" convention for the
default name of an executable file if no -o option is given to it. I'm
currently of the mind that this naming convention is completely outdated
and not worth propagating in llvmc. llvmc would require a -o option or
you get an error.
I don't see anything wrong with having the default for the "-o" option
be "a.out"... If it were up to me, I would keep it.
I don't see anything wrong with having the default for the "-o" option
be "a.out"... If it were up to me, I would keep it.
Well, the issue for me is that "a.out" in LLVM land can be a shell
script and even if its a native executable, it probably isn't in the
ancient a.out format.
perhaps we need "some" default, but not "a.out"?
or, are you saying this default is so ingrained into developer's minds
that it just needs to be supported regardless of the fact that it no
longer makes sense?
I'm just saying that the default makes sense because it has traditionally been
the default in other compilers that users will be familiar with. Whether or
not we're still using the a.out executable format doesn't directly pertain to
the question of what the default name should be -- I don't think any ELF
linkers name their default output "elf", for example.
In this situation where we're not offering a vast improvement upon the
traditional state of affairs, we should go with what people expect.