Re: [PATCH] gdbadmin/ss/gdb_ari.sh obsolete->regression

["Ulrich Weigand" <uweigand@de.ibm.com>]

Pierre Muller wrote:

>   I would tend to agree with you,
> but it's just more work...
>   So if anyone is volunteering, fine with me,
> I doubt that I will have time to eradicate all
> those unused macros from sources.
> 
> Pierre
> 
> PS: I also think that it would probably be good to
> leave the unused functions/macros in at least one
> GDB release, as people using GDB source code
> might still be using such obsolete code,
> especially if it was not deprecate'd.

It seems me the point of the ARI "regression" category
is to help catch adding back code patterns that we've
already managed to get rid of.  This makes perfect sense
for code patterns that *can* get added back accidentally
(e.g. coding style violations etc.).

However, some code patterns that are currently listed
as ARI regressions *cannot* be added back accidentally
anyway; for example, there is an ARI regression check
for use of the read_register function -- this function
is not even defined anywhere in the GDB source tree
today, so any attempt to call it would already cause
compile/link time failures.

I'd argue that ARI tests like those ought to be simply
deleted; they serve no further purpose.


B.t.w. as you're clearing up ARI results anyway (b.t.w.
thank you for your efforts in this area!), there's a
number of ARI tests where all remaining hits are in
fact false positives:

In the Obsolete category:

supply_register  (this function no longer exists; ARI
finds local struct members with the same name as false
positive results)

inside_entry_func  (this is now a static routine; the
original reason for this ARI entry no longer applies)

In the Deprecate category:

SP_REGNUM, PC_REGNUM, FP0_REGNUM   (these are not defined
anywhere any more; ARI finds local definitions with 
extended names like E_SP_REGNUM etc.)

In the Legacy category:

legacy_name  (this doesn't exist any more; ARI finds a
local variable with the same name)

I think these should also be simply deleted ...


Bye,
Ulrich


-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com