Re: [PATCH] PR c++/28986 failure to diagnose overflow for binary operators

["Manuel LÃpez-IbÃÃez" <lopezibanez at gmail dot com>]

On 06 Jan 2007 19:10:48 +0100, Gabriel Dos Reis
<gdr@integrable-solutions.net> wrote:
> | One particular issue that I am not sure about is whether initialisers
> | for static variables need to be constant in C++ as they do in C. The
> | relevant hunk in the diff between C and C++ testcases is:
> |
> | -/* But this expression does need to be constant.  */
> | +/* But this expression does need to be constant (in C++ ???).  */
>
> No, C++ does not have that restriction.
>
> (there is a new category of constant expressions in the pipeline, but
> it is too early to devise a testcase for that right now).
>
> The patch is OK, modulo removal of the C-only restriction.

I have updated (and bootstrapped and regression tested again) the
patch to reflect this. Also, I have removed the testcase
g++.dg/conversion/nullptr1.C since it is already included and
thoroughly tested in g++.dg/warn/overflow-warn-*.C

Another issue is that g++.dg/warn/overflow-warn-*.C testcases are a
modified copy of the corresponding gcc.dg/overflow-warn-*.c. Should I
use "svn copy" and modify each of them to keep the history? Or should
I just create them as new files to avoid any trouble?

Cheers,

Manuel.

Attachment: c-c++-overflow-testsuite.diff
Description: Text document

Attachment: c++-overflow-warning.diff
Description: Text document