format directives should be more forgiving to negative arguments

[James Kalenius]

(format t "~vd" -100 -100) signals an error. This is not the case on other implementations and isn't really supported by the standard, which just says that at least that many columns are printed. In ansi-cl-tests not signalling an error is part of test format.d.29.

The rest of the code as written works fine with negative numbers, so the error call (in format-write-field) just has to be deleted.

Exists in my distro's version (1.2.8) and in github master.

[Stas Boukarev (stassats)]

I don't think anyone would intentionally supply negative arguments, so it's fair to not accept them.

[Christophe Rhodes (csr21-cantab)]

Stas Boukarev <email address hidden> writes:

> I don't think anyone would intentionally supply negative arguments, so
> it's fair to not accept them.

For what it's worth, I'm not totally convinced by this. If the column
width is being computed programmatically, you could easily end up with a
situation where you subtract one thing from another. If the user is
required never to supply a negative argument, they will have to decorate
their code with (max 0 ...) everywhere -- and they may not detect the
problem until their code is in the field and exposed to the wider set of
inputs that deployment brings.

Christophe

[Stas Boukarev (stassats)]

That's true for many things, we do not promote (subseq sequence -10) to (subseq sequence 0)