Performance samples for LibreOffice ...

[Jan Hubicka]

> On 09/20/2014 04:39 AM, Jan Hubicka wrote:
> >/usr/include/X11/Xlib.h:948:3: warning: type ‘struct XAnyEvent’ violates one definition rule [-Wodr]
> >  } XAnyEvent;
> >    ^
> >/usr/include/X11/Xlib.h:948:3: note: a different type is defined in another translation unit
> >  } XAnyEvent;
> >    ^
> >/usr/include/X11/Xlib.h:947:9: note: the first difference of corresponding definitions is field ‘window’
> >   Window window; /* window on which event was requested in event mask */
> >          ^
> >/usr/include/X11/Xlib.h:947:9: note: a field of same name but different type is defined in another translation unit
> >   Window window; /* window on which event was requested in event mask */
> >          ^
> >/usr/include/X11/X.h:96:13: note: type ‘XLIB_Window’ should match type ‘Window’
> >  typedef XID Window;
> >              ^
> >/usr/include/X11/X.h:96:13: note: the incompatible type is defined here
> >  typedef XID Window;
> >              ^
> 
> ...so unwinding of typedefs (a la, "note: type 'XID' is defined
> here") could be helpful in tracking this down.

Typedefs are transparent for type mangling, so I do not thing you get warnings
on those at all.  Here I think the problem is that someone gets idea to #define
Window ZLIB_Window perhaps here
http://dev-builds.libreoffice.org/lcov_reports/master~2014-04-11_00.55.10/extensions/source/plugin/inc/plugin/unx/plugcon.hxx.gcov.html

I am not sure if that can be warned about better. One problem is that I
probably should name the incompatible type that I don't because the unified way
I output warnings.

Concerning include stack, I am honestly not sure why it is not output - will
try to track it down.  Also do not know why the ^ markers tends to be output on
beggining of the line in some cases.  I also plan to store actual compilation
unit name in case single source is used to be built with multiple #defines, so
that would give at least the top of inline stack.

Honza
> 
> Stephan