Re: [PATCH] posix_builtins: allow exporting a reaonly

[Martijn Dekker <martijn@inlv.org>]

Op 18-04-18 om 22:07 schreef Bart Schaefer:
> On Wed, Apr 18, 2018 at 12:58 PM, Martijn Dekker <martijn@inlv.org> wrote:
>> POSIX_BUILTINS incorrectly prohibits exporting a readonly variable. All
>> other POSIX shells allow this and there is nothing in the POSIX text[*] that
>> says it's not allowed. The attached patch fixes this.
> 
> I hope it's really that simple -- see thread about typeset -T and
> read-only and resulting confusion, e.g. zsh-workers/42061 and
> references.

I can't really find anything in that thread or its references that seems 
relevant.

Exporting readonly variables was already possible with POSIXBUILTINS 
off, so I don't see how it could harm anything to allow it with 
POSIXBUILTINS on.

Here is some more context than was provided by the patch. Note that all 
of this is only executed if POSIXBUILTINS is on.

     if (isset(POSIXBUILTINS)) {
         /*
          * Stricter rules about retaining readonly attribute in this case.
          */
         if ((on & (PM_READONLY|PM_EXPORTED)) &&
             (!usepm || (pm->node.flags & PM_UNSET)) &&
             !ASG_VALUEP(asg))
             on |= PM_UNSET;
         else if (usepm && (pm->node.flags & PM_READONLY) &&
                  !(on & PM_READONLY) && !(on & PM_EXPORTED)) {
             zerr("read-only variable: %s", pm->node.nam);
             return NULL;
         }
     }

As the comment says, that block applies stricter rules about retaining 
the readonly attribute for POSIXBUILTINS. I merely removed the 
restriction that readonly variables cannot be exported by adding:

	 && !(on & PM_EXPORTED)

Note that the restriction was already inconsistent: this code did not 
block unset readonly variables from being given the export flag.

FWIW, the test suite passes cleanly with this patch.

Thanks,

- M.