-O2 in Debug build?

[Alexander Neundorf]

On Friday 18 July 2008, Thiago Macieira wrote:
> Alexander Neundorf wrote:
> >On Friday 18 July 2008, Andreas Pakulat wrote:
> >> Hi,
> >>
> >> Whats the reasoning behind using -O2 in the Debug CXX flags? A debug
> >> build shouldn't use _any_ optimizations IMHO.
> >
> >I think there were lengthy discussion about the flags for the different
> > build types. I guess this was the result.
> >/me hasn't slept a lot this week, so I don't feel like searching for
> > this now
>
> The reason is that we split the build types in (for gcc):
>
> Release
> 	optimised for speed, qDebug/kDebug turned off, no debug symbols
> Release with debug info
> 	optimised for speed, debugging symbols on (-g)
> Debug
> 	optimised but debuggable, debugging on (-g)
> 	(-fno-reorder-blocks -fno-schedule-insns -fno-inline)
> DebugFull
> 	no optimisation, full debugging on (-g3)
> Profile
> 	DebugFull + -ftest-coverage -fprofile-arcs
>
> It is expected that the "Debug" build type be still debuggable with gdb
> without going all over the place, but still produce better performance.
> It's also important to note that gcc cannot detect all warning conditions
> unless the optimiser is active.
>
> I'd say that "debug" is recommended for kdelibs for people developing KDE
> applications, since oftentimes they may have to go into kdelibs
> functions. Their own code should be build under debugfull while still in
> development, but should be built a few times in release before a release.

Thanks for the explanation :-)
Can you put that on techbase, somewhere related to cmake/building ?

Alex