Re: [lmi] [lmi-commits] master e41c2a5: Forestall gcc-4.9.2 external attribute warnings, improved

[Greg Chicares]

On 2017-03-11 17:07, Vadim Zeitlin wrote:
> On Sat, 11 Mar 2017 10:31:55 -0500 (EST) Greg Chicares <address@hidden> wrote:
> 
> GC> branch: master
> GC> commit e41c2a5f8a7a17130afac6944299e9bfc05b0537
> GC> Author: Gregory W. Chicares <address@hidden>
> GC> Commit: Gregory W. Chicares <address@hidden>
> GC> 
> GC>     Forestall gcc-4.9.2 external attribute warnings, improved
> GC>     
> GC>     Reverted the 20151221T0138Z change for 'wx_new_test.cpp' only, and
> GC>     reworked that change in a better way. The unit test explicitly
> GC>     shared-library attribute, inducing a warning that was suppressed with
> GC>     a #pragma. Now, it is simply not endowed with any such attribute, so
> GC>     that no warning arises or requires suppression.
> 
>  This is not really a comment about this commit itself, which is a change
> for the better, of course, but I wonder why don't we use wx_new DLL in this
> test. It would seem to be more useful to check if the actual DLL works
> rather than using its code in a different way, from the one in which it's
> used in the main program, in the test.

That's probably a good idea. It would lose the advantage that the present
test compiles the wx_new code with more warnings enabled; however...
 - I haven't verified that the wx_new shared library actually is
   compiled without warnings that are incompatible with wx, and
 - if it is, then perhaps it shouldn't be; and of course
 - that's not a really strong objection, even if valid.

But my main thrust right now is to improve the shared-library macros,
and I had this pending patch (to remove some #pragma abuse) that was in
the way, so I cleared it first. That's also why I didn't at this time
address the comment pointing out that the unit test hardly tests
anything.