[PATCH] a little optimization of KJS GarbageCollector

Lubos Lunak l.lunak at suse.cz
Thu Dec 1 15:49:08 GMT 2005

On Thursday 01 December 2005 16:24, Dirk Mueller wrote:
> On Thursday 01 December 2005 16:09, David Faure wrote:
> > , for
> > reliable step-by-step debugging in gdb and backtraces.
> for reliable backtraces -fno-inline is enough :)

 Backtraces maybe. Step-by-step debugging is a pain without -O0.

> > So, like everyone
> > who compiles kdelibs with debug enabled, I'm among the developers who
> > benefit from an optimization like the one this thread is about;
> Note that I don't object to the patch getting included at all (except that
> it makes us differ from safari tree for no good reason). I'm just pointing
> out that gcc's output from -O0 code is so utterly horrible and slow that
> the least of your worries is this particular patch.

 And with -O2 gcc itself is so utterly horrible and slow ... . Not that I care 
that much myself, but not everybody has a fast machine and an icecream 
cluster that even the fast machine can't saturate.

> > and this
> > has nothing to do with -Wall does it? --enable-debug enables warnings too
> > iirc, but how is that related?
> -Wall does not work properly without -O2. With -O0 ALL code and dataflow
> analysis steps are disabled, and therefore warnings which use this
> information to trigger at all don't work either.
> > What do you expect? That everyone has two versions of kdelibs compiled,
> > one with debug for debugging and one without debug for more performance
> > and seeing all warnings? I don't think this is realistic.
> I just expect people to use realistic compiler flags.

 According to what you've written above such thing doesn't exist with gcc in 
practice. Which means we can skip this and just check if the patch makes 
sense in practice or not.

Lubos Lunak
KDE developer
SuSE CR, s.r.o.  e-mail: l.lunak at suse.cz , l.lunak at kde.org
Drahobejlova 27  tel: +420 2 9654 2373
190 00 Praha 9   fax: +420 2 9654 2374
Czech Republic   http://www.suse.cz/

More information about the kde-core-devel mailing list