Re: PR gdb/856

[Joel Brobecker <brobecker@adacore.com>]

:REVIEWMAIL:

> The PR suggests that parse_exp_1 ought to take a sal instead, so this
> is what I've implemented.  I changed any caller with access to a
> relevant sal or PC to use that; otherwise I changed the code to use
> either a sal constructed from the block's location, or an empty sal.

I am not sure about changing the block argument into a sal, and would
argue that changing it into a PC would be slightly better.  Indeed,
apart from the macro-scoping that uses a SAL, all the parse routines
only really need a PC. Using a SAL would force some of the callers
that don't already have one at hand to compute it.

A nice corolary is that it would simplify your patch quite a bit
by getting rid of the need to: replace expression_context_pc by
expression_context_sal; and define a new function empty_sal.

Thoughts? Andrew Cagney and Jim Blandy both thought using a SAL
would be a good idea, so the SAL idea does have some weight, but...

-- 
Joel