R devel - Jul 2004 - Segfault with options() (PR#7078)

This is completely undocumented use of options, but it does cause a 
segfault with r-patched and r-devel on Linux:

options(list('digits', 'scipen'))

Strangely enough, 

options(list('digits'))

and 

options(list(digits = 10, scipen = 2))

work as might be expected.

Deepayan

Thank you, Deepayan:
>>>>> "Deepayan" == Deepayan Sarkar
<deepayan@stat.wisc.edu>
>>>>>     on Mon, 12 Jul 2004 05:46:08 +0200 (CEST) writes:
    Deepayan> This is completely undocumented use of options,
    Deepayan> but it does cause a segfault with r-patched and
    Deepayan> r-devel on Linux:

    Deepayan> options(list('digits', 'scipen'))

    Deepayan> Strangely enough,

    Deepayan> options(list('digits'))

    Deepayan> and

    Deepayan> options(list(digits = 10, scipen = 2))

    Deepayan> work as might be expected.

I think I've seen where and how the problem occurs and am
currently testing an easy patch.

That will however give an error message for all calls 
   options( mylist )
when mylist is a list that has  
  length(names(mylist)) !=  length(mylist)

You weren't making a "feature request" in the other direction 
where you? 

Regards,
Martin Maechler

On Monday 12 July 2004 11:37, Martin Maechler wrote:
> Thank you, Deepayan:
> >>>>> "Deepayan" == Deepayan Sarkar
<deepayan@stat.wisc.edu>
> >>>>>     on Mon, 12 Jul 2004 05:46:08 +0200 (CEST) writes:
>
>     Deepayan> This is completely undocumented use of options,
>     Deepayan> but it does cause a segfault with r-patched and
>     Deepayan> r-devel on Linux:
>
>     Deepayan> options(list('digits', 'scipen'))
>
>     Deepayan> Strangely enough,
>
>     Deepayan> options(list('digits'))
>
>     Deepayan> and
>
>     Deepayan> options(list(digits = 10, scipen = 2))
>
>     Deepayan> work as might be expected.
>
> I think I've seen where and how the problem occurs and am
> currently testing an easy patch.
>
> That will however give an error message for all calls
>    options( mylist )
> when mylist is a list that has
>   length(names(mylist)) !=  length(mylist)
>
> You weren't making a "feature request" in the other direction
> where you?
No. I'm playing around with the possibility of having similarly behaved 
functions like lattice.options() and lattice.getOption(), and I was 
checking the behaviour of options() when I encountered the segfault.

Should options( mylist ) work at all? It's not documented, at least not 
in ?options. Is that a documentation oversight?

Deepayan

On Mon, 12 Jul 2004 18:54:22 +0200 (CEST), deepayan@stat.wisc.edu
wrote :
>Should options( mylist ) work at all?
Yes, that's used fairly often in the pattern

 save <- options( <something new> )
 on.exit(options(save))
>  It's not documented, at least not 
>in ?options.  Is that a documentation oversight?
Looks like it.  ?par, which uses similar syntax, documents ... as

     ...: arguments in 'tag = value' form, or a list of tagged values.
          The tags must come from the graphical parameters described
          below.

as opposed to ?options, which leaves off the "or a list" part.

Duncan Murdoch