Re: Logging --version

[Akim Demaille]

>>>>> "Alexandre" == Alexandre Oliva <address@hidden> writes:

Alexandre> On Apr 4, 2001, Akim Demaille <address@hidden> wrote:
>> This is what I fear.  What could happen?

Alexandre> Hmm...  Or getting a file named `--version' created in the
Alexandre> current directory, when testing for touch.  Or getting this
Alexandre> blank file printed, when testing for lpr later on?  Or
Alexandre> getting a shell script named --version, that happens to be
Alexandre> in your PATH, to be run, when testing for some particular
Alexandre> shell?  Or getting an error message in your mail box saying
Alexandre> that there's no mailbox named --version, when testing for
Alexandre> mail delivery programs?  I'm sure someone could come up
Alexandre> with more terrifying examples.

Wow, that was quite a thrill :)  OK, I'm scared :)

But in fact, I agree we can build examples like this, my question was
more whether it is *likely* to happen (with `likely' referring to `not
being built to this end'[1]).

Alexandre> I'd rather special-case --version so that we only do it
Alexandre> when we know it's a reasonable safe thing to do.

>> Say, compilers?

Alexandre> Yep.  Perhaps tools in general (as in AC_CHECK/PATH_TOOL,
Alexandre> as opposed to AC_CHECK/PATH_PROG).

I like this!


[1] Are there really programs out there that don't choke on
--sldkljfjlsdf but try to process it?