Re: [PATCH v3 3/5] GDB: Only make data actually retrieved into value history available

[Tom Tromey <tom@tromey.com>]

>>>>> "Maciej" == Maciej W Rozycki <macro@embecosm.com> writes:

Maciej> While it makes sense to allow accessing out-of-bounds elements in the 
Maciej> debuggee and see whatever there might happen to be there in memory (we 
Maciej> are a debugger and not a programming rules enforcement facility and we 
Maciej> want to make people's life easier in chasing bugs), e.g.:

Thank you for the patch.

Maciej> Index: src/gdb/defs.h
Maciej> ===================================================================
Maciej> --- src.orig/gdb/defs.h
Maciej> +++ src/gdb/defs.h
Maciej> @@ -469,6 +469,10 @@ enum val_prettyformat
Maciej>  #define	LONG_MAX ((long)(ULONG_MAX >> 1))   /* 0x7FFFFFFF for 32-bits */
Maciej>  #endif
 
Maciej> +#if !defined (LONG_MIN)                     /* 0x80000000 for 32-bits */
Maciej> +#define	LONG_MIN ((long) (~(long) 0 ^ LONG_MAX))
Maciej> +#endif

I wasn't aware of this code being here.

For LONG_*, I tend to think we don't need it.  <limits.h> should provide
it, and gnulib seems to think there aren't any systems without this:

https://www.gnu.org/software/gnulib/manual/html_node/limits_002eh.html

 
Maciej> +#if !defined (LONGEST_MIN)                 /* 0x8000000000000000 for 64-bits */
Maciej> +#define	LONGEST_MIN ((LONGEST) (~(LONGEST) 0 ^ LONGEST_MAX))
Maciej> +#endif

This one probably belongs in gdbsupport/common-types.h.

We should clean up this part of defs.h but I can do that.

Maciej> @@ -1950,6 +1955,15 @@ record_latest_value (struct value *val)
Maciej>       a value on the value history never changes.  */
Maciej>    if (value_lazy (val))
Maciej>      value_fetch_lazy (val);
Maciej> +
Maciej> +  /* Don't pretend we have anything available there in the history beyond
Maciej> +     the boundaries of the value recorded.  It's not like inferior memory
Maciej> +     where there is actual stuff underneath.  */
Maciej> +  ULONGEST length = value_enclosing_type (val)->length ();
Maciej> +  mark_value_bits_unavailable (val, LONGEST_MIN, 0 ^ LONGEST_MIN);
Maciej> +  mark_value_bits_unavailable (val, length * TARGET_CHAR_BIT,
Maciej> +			       LONGEST_MAX - length * TARGET_CHAR_BIT);

I'm struggling to understand why this is needed.
Like it seems to me that a value should always report that bytes outside
of its content range are unavailable.

Tom