GNU bug report logs - #29617
seq: `seq 1 --help' doesn't give help

Previous Next

Full log

View this message in rfc822 format

From: Eric Blake <eblake <at> redhat.com>
To: chadweisman <at> nym.hush.com, 29617 <at> debbugs.gnu.org
Subject: bug#29617: `seq 1 --help' doesn't give help
Date: Fri, 8 Dec 2017 12:55:04 -0600

[Message part 1 (text/plain, inline)]

On 12/08/2017 11:38 AM, chadweisman <at> nym.hush.com wrote:
> Hello,
> 
> I am using coreutils version 8.27 on Fedora, and I don't see this fixed in 8.28's NEWS.
> 
> $ seq 1 --help
> seq: invalid floating point argument: ‘--help’
> Try 'seq --help' for more information.

Interesting bug!

> 
> We should be able to put the options anywhere and not necessarily before any arguments.

Yes, when possible.

>  And even if not (e.g. POSIX conformance overrides,)

POSIX does say you have to write 'foo -- --help' if you want to
guarantee that --help is treated as a literal argument rather than
option, but it also says that the moment you specify '--help' (or any
other parameter starting with two dashes without the -- end-of-options
parameter), you are already in undefined territory.  So we can do
whatever we want when encountering '--help' - which is part of the
reason WHY the GNU project prefers making 'foo args --help' print help
output where possible.

> --help should be handled specially to conform to the GNU coding standards. [1]

Yes.

But the reason that it fails is because we use getopt_long(...,
"+f:s:w") - where the leading '+' specifically requests that we NOT
allow option reordering.  Why? Because 'seq' is MOST useful if it can
parse negative numbers easily.  We deemed it more important to support
'seq 2 -1 1' without requiring the user to write 'seq -- 2 -1 1' - but
in doing so, it also means that we can't reorder options, so any obvious
non-option (like '1' in your example) makes all other parameters
non-options (including '--help' in your example).

It might be possible to do a two-pass parse over argv: one that looks
just for --help (and where treating -1 as an option is a no-op), and the
second that actually parses things in order now that it knows --help is
not present.  But that's a lot of code to add for a corner case, so I
won't be writing the patch; but I also won't turn it down if someone
else wants to write the patch.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org

[signature.asc (application/pgp-signature, attachment)]

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.