Re: mdocml: Finally use __progname, err(3) and warn(3).

Re: mdocml: Finally use __progname, err(3) and warn(3).

[Joerg Sonnenberger]

On Sun, Oct 11, 2015 at 04:13:25PM -0500, schwarze@mdocml.bsd.lv wrote:
> Finally use __progname, err(3) and warn(3).
> That's more readable and less error-prone than fumbling around
> with argv[0], fprintf(3), strerror(3), perror(3), and exit(3).
> 
> It's a bad idea to boycott good interfaces merely because standards
> committees ignore them.  Instead, let's provide compatibility modules 
> for archaic systems (like commercial Solaris) that still don't have
> them.  The compat module has an UCB Copyright (c) 1993...

No, the reason to avoid it is because __progname is in the
implementation namespace and should be avoided as such. That's why in
NetBSD we introduced setprogname() and getprogname(). On systems without
__progname, they provided a clean abstraction without messing with the
implementation namespace. On platforms with __progname, but without
getprogname, the latter is trivially implemented as returning __progname
and setprogname is a nop.

Joerg
--
 To unsubscribe send an email to discuss+unsubscribe@mdocml.bsd.lv

Re: mdocml: Finally use __progname, err(3) and warn(3).

[Ingo Schwarze]

Hi Joerg,

Joerg Sonnenberger wrote on Fri, Nov 06, 2015 at 01:28:57PM +0100:
> On Sun, Oct 11, 2015 at 04:13:25PM -0500, schwarze@mdocml.bsd.lv wrote:

>> Finally use __progname, err(3) and warn(3).
>> That's more readable and less error-prone than fumbling around
>> with argv[0], fprintf(3), strerror(3), perror(3), and exit(3).
>> 
>> It's a bad idea to boycott good interfaces merely because standards
>> committees ignore them.  Instead, let's provide compatibility modules 
>> for archaic systems (like commercial Solaris) that still don't have
>> them.  The compat module has an UCB Copyright (c) 1993...

> No, the reason to avoid it is because __progname is in the
> implementation namespace and should be avoided as such.

My rant in the commit message was intended to address err(3),
not __progname.  But i didn't make that clear.

> That's why in NetBSD we introduced setprogname() and getprogname().
> On systems without __progname, they provided a clean abstraction
> without messing with the implementation namespace.
> On platforms with __progname, but without getprogname, the latter
> is trivially implemented as returning __progname
> and setprogname is a nop.

I just committed a patch to switch from __progname to getprogname(3),
thanks for the suggestion.

I didn't bother with platforms having __progname but not getprogname(3);
those can use argv[0] just like platforms having neither.

Yours,
  Ingo
--
 To unsubscribe send an email to discuss+unsubscribe@mdocml.bsd.lv

Re: mdocml: Finally use __progname, err(3) and warn(3).

[Joerg Sonnenberger]

On Fri, Nov 06, 2015 at 05:43:57PM +0100, Ingo Schwarze wrote:
> I didn't bother with platforms having __progname but not getprogname(3);
> those can use argv[0] just like platforms having neither.

That's fine. The note was primarily about addressing old rants from Theo
about getprogname() :) 

Joerg
--
 To unsubscribe send an email to discuss+unsubscribe@mdocml.bsd.lv

Re: mdocml: Finally use __progname, err(3) and warn(3).

[Ingo Schwarze]

Hi Joerg,

Joerg Sonnenberger wrote on Fri, Nov 06, 2015 at 05:47:30PM +0100:
> On Fri, Nov 06, 2015 at 05:43:57PM +0100, Ingo Schwarze wrote:

>> I didn't bother with platforms having __progname but not getprogname(3);
>> those can use argv[0] just like platforms having neither.

> That's fine. The note was primarily about addressing old rants from
> Theo about getprogname() :) 

I seem unable to find those, but i understand that he encouraged
import of getprogname(3) into OpenBSD libc and that he later
said it "has the potential to be more portable":

  http://marc.info/?l=openbsd-cvs&m=137003410606499&w=2
  http://marc.info/?l=openbsd-tech&m=139882367211410&w=2

Yours,
  Ingo
--
 To unsubscribe send an email to discuss+unsubscribe@mdocml.bsd.lv