Re: [hackers] [sbase][PATCH] basename, dirname, printf: recognise -- and fail if options are used.

[Michael Forney]

On 12/27/16, Evan Gates  wrote:
>> I respect your opinion and have to admit that this is not an easy
>> discussion. Let's wait for some feedback and see what the others think
>> about it.
>
> I side with Mattias on this one. Accepting -- even in utilities that
> don't have flags has the benefits of 1) complying with POSIX so we can
> be a drop in replacement 2) consistency with all the other tools.

Actually, POSIX talks about '--' in guideline 10 of the Utility Syntax
Guidelines[0]

Many tools, for example cp, say

  The cp utility shall conform to XBD Utility Syntax Guidelines .

But this is not mentioned for printf. So, the current implementation
does comply with POSIX (at least in this respect).

However, I did run into an issue with musl-cross-make which was
passing '--' to printf. This has been fixed[1] in the latest release
though. Are there many other places where scripts use -- with printf?
Can we fix those instead?

[0] 
http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap12.html#tag_12_02
[1] 
https://github.com/richfelker/musl-cross-make/commit/3ac08b98c1c947043df489a760135ccb7f8b783d

Re: [hackers] [sbase][PATCH] basename, dirname, printf: recognise -- and fail if options are used.

[Evan Gates]

On Tue, Dec 27, 2016 at 7:05 AM, Laslo Hunhold  wrote:
> On Tue, 27 Dec 2016 15:48:56 +0100
> Mattias Andrée  wrote:
>
> Hey Mattias,
>
>> Okay, I personally do not agree with this and see echo(1)
>> as an abomination, it treats any unrecognised flags as
>> strings, but if had debate on it, keep to want you agreed
>> to.

I agree echo is an abomination. Especially with POSIX saying things
like "If the first operand is -n, or if any of the operands contain a
 character, the results are implementation-defined." The
best thing we can do is not use echo in any scripts.

I think from a philosophical standpoint sbase echo should not include
-n, to be simpler and still POSIX compliant. However pragmatically
many scripts use echo -n so I think it makes sense to keep it.

>
> I respect your opinion and have to admit that this is not an easy
> discussion. Let's wait for some feedback and see what the others think
> about it.

I side with Mattias on this one. Accepting -- even in utilities that
don't have flags has the benefits of 1) complying with POSIX so we can
be a drop in replacement 2) consistency with all the other tools.

Re: [hackers] [sbase][PATCH] basename, dirname, printf: recognise -- and fail if options are used.

[Laslo Hunhold]

On Tue, 27 Dec 2016 15:48:56 +0100
Mattias Andrée  wrote:

Hey Mattias,

> Okay, I personally do not agree with this and see echo(1)
> as an abomination, it treats any unrecognised flags as
> strings, but if had debate on it, keep to want you agreed
> to.

I respect your opinion and have to admit that this is not an easy
discussion. Let's wait for some feedback and see what the others think
about it.

Cheers

Laslo

-- 
Laslo Hunhold 

Re: [hackers] [sbase][PATCH] basename, dirname, printf: recognise -- and fail if options are used.

[Laslo Hunhold]

On Tue, 27 Dec 2016 15:26:09 +0100
Mattias Andrée  wrote:

Hey Mattias,

> POSIX explicitly states that echo(1) shall treat “--”
> as a string operand and not as a delimiter. My guess
> is that this is for historical reason, much like the
> existence of echo(1) itself.

ah now I remember, thanks. You know, sbase takes the liberty of moving
away from Posix where consistency is a concern. In this case I think
our result from a long debate was to say: Well, we just treat "--" as a
normal string for tools that do not accept any flags.

For instance, if echo does not interpret --, why doesn't printf also
follow echo in this behaviour?

Cheers

Laslo

-- 
Laslo Hunhold 

Re: [hackers] [sbase][PATCH] basename, dirname, printf: recognise -- and fail if options are used.

[Mattias Andrée]

On Tue, 27 Dec 2016 15:21:11 +0100
Laslo Hunhold  wrote:

> On Tue, 27 Dec 2016 15:14:44 +0100
> Mattias Andrée  wrote:
> 
> Hey Mattias,
> 
> > POSIX says “--” should be supported unless stated
> > otherwise. I interpret that standard saying that this
> > also applies utilities that do not take any flags. And
> > to be on the safe side I think it is a good idea to
> > support “--” for two reasons: (1) existing scripts may
> > require it, and (2) if POSIX adds a flag to the utility
> > in the future we must support “--” now, otherwise
> > compatibility will be broken.  
> 
> so what about "echo --"?
> 
> Cheers
> 
> Laslo
> 

POSIX explicitly states that echo(1) shall treat “--”
as a string operand and not as a delimiter. My guess
is that this is for historical reason, much like the
existence of echo(1) itself.


pgpkMrgLX4Pr4.pgp
Description: OpenPGP digital signature

Re: [hackers] [sbase][PATCH] basename, dirname, printf: recognise -- and fail if options are used.

[Laslo Hunhold]

On Tue, 27 Dec 2016 15:14:44 +0100
Mattias Andrée  wrote:

Hey Mattias,

> POSIX says “--” should be supported unless stated otherwise.
> I interpret that standard saying that this also applies utilities
> that do not take any flags. And to be on the safe side I think
> it is a good idea to support “--” for two reasons: (1) existing
> scripts may require it, and (2) if POSIX adds a flag to the
> utility in the future we must support “--” now, otherwise
> compatibility will be broken.

so what about "echo --"?

Cheers

Laslo

-- 
Laslo Hunhold 

Re: [hackers] [sbase][PATCH] basename, dirname, printf: recognise -- and fail if options are used.

[Mattias Andrée]

On Tue, 27 Dec 2016 11:26:07 +0100
Laslo Hunhold  wrote:

> On Fri, 25 Mar 2016 21:31:37 +0100
> Mattias Andrée  wrote:
> 
> Hey Mattias,
> 
> >  basename.c |  5 -
> >  dirname.c  |  6 +-
> >  printf.c   | 12 
> >  3 files changed, 17 insertions(+), 6 deletions(-)  
> 
> I do not support this patch, as "--" only makes sense for
> tools that actually take flags.
> I pity that GNU basename(1), dirname(1) and printf(1)
> take flags, however, see great value in being able to do
> 
>   $ printf --
> 
> and get what you asked for instead of some error message
> 
>   printf: usage: printf [-v var] format [arguments]
> 
> Cheers
> 
> Laslo
> 

POSIX says “--” should be supported unless stated otherwise.
I interpret that standard saying that this also applies utilities
that do not take any flags. And to be on the safe side I think
it is a good idea to support “--” for two reasons: (1) existing
scripts may require it, and (2) if POSIX adds a flag to the
utility in the future we must support “--” now, otherwise
compatibility will be broken.


pgpNxFM3rLXhu.pgp
Description: OpenPGP digital signature