This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFC][02/19] Target FP: Simplify floatformat_from_type
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: simon dot marchi at polymtl dot ca (Simon Marchi)
- Cc: gdb-patches at sourceware dot org
- Date: Mon, 18 Sep 2017 13:49:51 +0200 (CEST)
- Subject: Re: [RFC][02/19] Target FP: Simplify floatformat_from_type
- Authentication-results: sourceware.org; auth=none
Simon Marchi wrote:
> On 2017-09-05 20:20, Ulrich Weigand wrote:
> > size_t len = bit / TARGET_CHAR_BIT;
> > - gdb_assert (len >= floatformat_totalsize_bytes (floatformats[0]));
> > - gdb_assert (len >= floatformat_totalsize_bytes (floatformats[1]));
> > + gdb_assert (len * TARGET_CHAR_BIT >= floatformat->totalsize);
>
> That looks funny now. Is there a reason not to do
>
> gdb_assert (bit >= floatformat->totalsize);
>
> directly?
Well, that would use different rounding ... Of course, this only
matters when floatformat->totalsize is not a multiple of TARGET_CHAR_BIT,
which doesn't happen for any of the floatformats GDB supports, so it
is probably moot. Maybe we should instead use an assert to verify that
floatformat->totalsize is in fact a multiple of TARGET_CHAR_BIT?
Bye,
Ulrich
--
Dr. Ulrich Weigand
GNU/Linux compilers and toolchain
Ulrich.Weigand@de.ibm.com