Re: [PATCH v2] gdb: split array and string limiting options

["Maciej W. Rozycki" <macro@embecosm.com>]

On Wed, 26 Jan 2022, Simon Sobisch wrote:

> > > > > To do so the new parameter cannot be directly used as it is in the
> > > > > patch
> > > > > but must be fowarded to a new function or duplicated.
> > > > > 
> > > > > As an example this would mean instead of :
> > > > > 
> > > > > --- src.orig/gdb/c-valprint.c
> > > > > +++ src/gdb/c-valprint.c
> > > > > @@ -271,7 +271,7 @@ c_value_print_array (struct value *val,
> > > > > 
> > > > >       for (temp_len = 0;
> > > > >    	   (temp_len < len
> > > > > -	    && temp_len < options->print_max
> > > > > +	    && temp_len < options->print_max_chars
> > > > >    	    && extract_unsigned_integer (valaddr + temp_len * eltlen,
> > > > >    					 eltlen, byte	_order) != 0);
> > > > >    		   ++temp_len)
> > > > > 
> > > > > have the following
> > > > > 
> > > > > 
> > > > > +   const int print_max_chars = (options->print_max_chars != -1) ?
> > > > > options->print_max_chars : options->print_max;
> > > > >       for (temp_len = 0;
> > > > >    	   (temp_len < len
> > > > > -	    && temp_len < options->print_max
> > > > > +	    && temp_len < print_max_chars
> > > > >    	    && extract_unsigned_integer (valaddr + temp_len * eltlen,
> > > > >    					 eltlen, byte_order) != 0);
> > > > >    		   ++temp_len)
> > > > 
> 
> Proposal 1 - Should work, shouldn't it?
> 
> >   If you want to keep backwards compatibility, then no particular value for
> > `set print elements' can be used to drive `set print characters' I'm
> > afraid.  Someone may have used it somewhere.
> 
> What do you mean with that? If the default is whatever print elements is it
> should be compatible, shouldn't it?

 Right, I got this mixed up as it's the new `set print characters' setting 
that will drive the default rather than the existing `set print elements' 
one, the semantics of which will remain unchanged for non-character 
entities.  So e.g.:

(gdb) set print characters elements

(which will also be the default) will make `set print elements' drive the 
number of characters printed (and the `elements' syntactic argument can 
either set the limit to a reserved value such as `-1' as you suggested or 
drive a separate internal boolean flag).  While if we use any other value, 
then the character limit will get separated from the `set print elements' 
setting.

 So it seems good to me and probably the least interface complication for 
users and the least maintenance burden to carry along.  Thank you for your 
persistence!

 Any other input, anyone?

  Maciej