Re: [Tsp-devel] TSP0.8.3 & tsb_bb_provider & tsp_ascii_writer

[Eric Noulard]

2010/5/4 David <address@hidden>:
> Hello!
>
> I have a problem using tsp_bb_provider with consumers gdisp and
> tsp_ascii_writer.
> Everything fine to read scalar value but problem reading a specific array
> value.
>
> On the provider:
> tsp_bb_provider BB 10
>
> On the consumer:
> tsp_ascii_writer -u rpc://server_ip -x test.dat
>
> with test.dat :
> symbol_array[2] 1
>
> The result is not symbol_array[2] (a scalar) but the entire array
> (symbol_array).

I have to check, but I think 0.8.3 cannot do that, in fact the
tsp_ascii_writer and tsp_gdisp do not handle "single element array
subscription"
but the core TSP do.

> With gdisp, I have the same problem (same unreadable graph for symbol_array
> or symbol_array[x]).
>
> The blackboard is created with TSP 0.8.3.
> Same probleme with a tsp_bb_provider and consumer from CVS source.

This is weird I thought we did at least modify tsp_ascii_writer to handle this
particular case. I'll check and keep you informed.

May be I'm wrong and/or the tsp_bb_provider does not handle the
request properly...

The [list of ] "TSP_sample_symbol_info_t" structure found in request sample
contains (see tsp_rpc.x):

/**
   * The dimension of the symbol (array size).
   * <ul>
   *   <li> 1 for scalar </li>
   *   <li> more than 1 for arrays. </li>
   * </ul>
   */
  unsigned int dimension;
  /**
   * The offset of the first requested element
   * if the symbol is an array.
   * Only used if dimension > 1.
   */
  unsigned int offset;
  /**
   * The number of requested element beginning at offset.
   * 0 means all element requested beginning at offset.
   * Only used if dimension > 1.
   */
  unsigned int nelem;

which is used to request "partial" values of an array.

> Any Idea?

Dive into the TSP code (of ascii writer) and look where
TSP_consumer_request_sample is called then go back (in the code)
where the "TSP_sample_symbol_info_list_t" is built in order to check
whether if the
appropriate request is built.

You may try ascii_writer with an alternate provider (stub,
ascii_reader) in order to see
if the "unimplemented" feature is on the provider side.

-- 
Erk
Membre de l'April - « promouvoir et défendre le logiciel libre » -
http://www.april.org