[major] struct array indexing in tip

[John W. Eaton]

On 20-Jan-2009, Jaroslav Hajek wrote:

| On Mon, Jan 19, 2009 at 3:45 AM, Nicholas Tung <ntung at ntung.com> wrote:
| > in revision
| > hg tip
| > changeset:   8534:0eb83938c8bc
| >
| > struct indexing is messed up. The following code should simply add to the
| > value of c at 2, but creates some weird cell array...
| > octave:1> a.c = [1, 2, 3]
| > a =
| > {
| >   c =
| >      1   2   3
| > }
| > octave:2> e = [a, a]
| > e =
| > {
| >   1x2 struct array containing the fields:
| >     c
| > }
| > octave:3> e(1).c(2)
| > ans =  2
| > octave:4> e(1).c(2) += 1
| > e =
| > {
| >   1x2 struct array containing the fields:
| >     c
| > }
| > octave:5> e.c
| > ans =
| > {
| >   [1,1] =
| >      1   2   3
| >   [1,2] =  3
| > }
| > ans =
| >    1   2   3
| >
| > Thanks in advance; please tell me when it's fixed.
| >
| > cheers,
| > Nicholas
| 
| Gosh, it took me about a day and half, and here goes:
| http://hg.savannah.gnu.org/hgweb/octave/rev/3d8a914c580e
| 
| Your bug seems fixed. Moreover, for instance, the following will now work:
| 
| [a(1:2).x] = deal (1, 3)
| [a(:).x(2)] = deal (5, 6)
| 
| I'm still not convinced all cases are caught, so I encourage everyone
| to experiment.
| In fact, I'd be surprised if it were really bug-free, but make check
| didn't catch anything.

Thanks for looking at this problem.

Maybe someone (other than Jaroslav) would like to generate some tests
so these problems don't resurface later?

jwe