Re: test failures with current mercurial sources (was Re: 3.0.1 release?)

[David Bateman]

John W. Eaton wrote:
> On  3-Apr-2008, David Bateman wrote:
>
> | Yes that might be a nice feature, but I think you'd then also need
> | 
> | warning ("clear", "Octave:deprecated-function");
> | 
> | so that the fact that the warning has already appeared might be
> | cleared..
>
> Couldn't the existing "warning on|off" do that job?
>
> | Will the fact that there are a large number of named warning
> | make things slower however? If so I think I'd prefer to forgo it..
>
> I could be wrong, but I think most warnings are only issued if some
> condition is met.
>
> For the case of using a deprecated function that would produce an
> unconditional warning, I don't think it matters that it is slower.  We
> want people to move away from the deprecated functions anyway, so it
> is just extra incentive if switching improves performance.
>
> In any case, when warning_with_id is called it does quite a bit of
> work and performs a linear search over all the elements of
> warning_options.  So yes, that could probably be improved.  The
> internal warning_state variable is a map, so I don't remember why it
> is searched linearly.  I guess because it is a structure array with
> elements "identifier" and "state" rather than a map from identifier to
> state.  Maybe we could change the internal representation to provide
> better performance while still providing the struct array format for
> compatibility, only generating it when requested.
>   
If we add lots of warnings with ids, then I think it would be a good
idea to use a hash to speed the identification of the warning in the map
rather than use a linear search.. Doesn't std::map include that in any case?
> The simple thing is to add the persistent variables to the deprecated
> functions as you suggested.  I don't object to that, but you asked
> whether there was another way.
>   
Well, I wasn't sure it was the best manner to add the warnings, and
hoped there might be a nicer way. I'll add them with a persistent
variable in a patch I'll send later.

D.

-- 
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary