Re: non-representable symbols in PE executables

["Jan Beulich" <JBeulich at suse dot com>]

>>> On 01.07.16 at 17:16, <JBeulich@suse.com> wrote:
>>>> On 01.07.16 at 13:08, <nickc@redhat.com> wrote:
>>> The most simple adjustment I could think of would be to associate
>>> symbols with the following section only if they're right at the start
>>> of that section. But of course I have no idea what other users of
>>> ld would break with a change to heuristics like this one.
>> 
>> Well, nobody has screamed, so please consider this patch approved.
> 
> Thanks, but before committing I thought it'd be a good idea to run
> the testsuite for some COFF and PE targets over this, and there are
> a number of failures, resulting from the added warning actually
> triggering on symbols with apparently negative (wrapped to huge
> positive) values. Removing the warning for the testsuite's sake
> doesn't feel right, and trying to ignore the warnings in the testsuite
> would also feel kind of hackish (albeit default_ld_simple_link already
> does so for another warning). Should I add a command line override
> to silence that warning, and make use of it for the affected tests?
> Do you have any other recommendation?

Or maybe, slightly reducing the intended effect of the change, limit
the warnings to the case where the value isn't representable when
considering it unsigned nor when considering it signed? I think that
would at least take care of all the testsuite failures (but it would
leave in place brokenness for certain consumers, including nm).

Jan