Re: [Tiger-devel] [RFC] Diferences between 'empty and set' and 'unset' variables

[Nicolas François]

On Fri, Sep 19, 2003 at 02:08:25AM +0200, Javier Fernández-Sanguino Peña wrote:
> On Tue, Sep 16, 2003 at 10:04:37AM +0200, Nicolas François wrote:
> > Hello,
> > 
> > Here are some patches and comments all related to tests of empty
> > variables. In some cases, I think a difference can be found between a
> > variable set to "" and an unset variable.
> 
> You might be right but it probably depends on context.

I have to confess that I mainly posted these patches to point out the
problem.


> > -- 08_initdefs_haveallvars_allow_empty_vars.patch ----------------------
> > What should haveallvars do with variables empty but set?
> 
> IMHO it should exit just like if it's undefined. I'm adding your patch but 
> providing a different error message for the other case. Notice that some
> tests will not work properly if a variable checked by haveallvars() turns 
> out to be empty.

You're right. And there would probably be no variable tested by
haveallvars which could be empty (not checked).

BTW, there were an error in my patch. I attach a new patch for
initdefs (initdefs_haveallvars.diff)



I also had another patch for initdefs:
  in function haveallcmds, the command is tested for execution by a '-f'.
  Souldn't '-x' used instead (or TESTEXEC)? Is it for systems that doesn't
  support '-x' (does it exist?)?
  A patch is attached in case you think it's worth being applied.
  (initdefs_exec.diff)
  I have the same question for config (TESTEXEC set to '-f'). I attach
  01_config_exec.patch just in case.


Regards,
-- 
Nekral

initdefs_haveallvars.diff
Description: Text document

initdefs_exec.diff
Description: Text document

01_config_exec.patch
Description: Text document