Hi,

attached to this mail, there is the output of scan with cppcheck. Cppcheck reports few issues about post and pre incrementation of iterators (post increment is faster) and const correctness,e.g:

class A
{
   public:
          bool IsOk() {return state;} // <-- cppcheck complains about a missing cont declaration
   bool state;
};

solution:
class A
{
   public:
          bool IsOk() const {return state;} 
   bool state;
};


One other main stylistic issue is that it is allowed to deallocate a NULL Pointer
e.g:

int *p = NULL;
delete p;

is totally legal.



I hope this is usefull for you?

Best regards

Ettl Martin




Hi,

of course, i can mail the whole output to you, if you wish? But i suggest to run it on yourself using the following commandline command:

Download cppcheck at http://sourceforge.net/projects/cppcheck/
cppcheck --enable=all [src-directory]

Best regards

Martin



-------- Original-Nachricht --------
> Datum: Mon, 22 Mar 2010 13:49:37 -0700 (PDT)
> Von: Sage Weil <sage@newdream.net>
> An: Martin Ettl <ettl.martin@gmx.de>
> CC: ceph-devel@lists.sourceforge.net
> Betreff: Re: [ceph-devel] [cppcheck] suggested a few performance improvements

> Hi Martin,
> 
> On Mon, 22 Mar 2010, Martin Ettl wrote:
> > i have checked the current git head with cppcheck ( a static code 
> > analysis tool, http://sourceforge.net/projects/cppcheck/). It is an 
> > allready very powerfull tool that makes some suggestions, how to improve
> > performance of the code.
> > 
> > Here is a snipped of the scan:
> > ....
> > [./src/mds/MDSMap.h:236]: (possible style) Pre-Incrementing variable 'p'
> is preferred to Post-Incrementing
> > [./src/mds/MDSMap.h:256]: (possible style) Pre-Incrementing variable 'p'
> is preferred to Post-Incrementing
> > [./src/mds/MDSMap.h:263]: (possible style) Pre-Incrementing variable 'p'
> is preferred to Post-Incrementing
> > [./src/mds/MDSMap.h:280]: (possible style) Pre-Incrementing variable 'p'
> is preferred to Post-Incrementing
> > [./src/mds/MDSMap.h:291]: (possible style) Pre-Incrementing variable 'p'
> is preferred to Post-Incrementing
> > [./src/include/LogEntry.h:100]: (possible style) Pre-Incrementing
> variable 'p' is preferred to Post-Incrementing
> > [./src/ceph.cc:317]: (possible style) Pre-Incrementing variable 'p' is
> preferred to Post-Incrementing
> > [./src/mds/MDSMap.h:247]: (possible style) Use failed.empty() instead of
> failed.size() to guarantee fast code.
> > ....
> > 
> > 
> > Please refer the attached patch that chanes a few of this suggestions.
> > What do you think about this? Does it boost the performance?
> 
> I don't think it'll affect performance in these cases, but stylistically 
> it's an improvement, so I'll apply it.
> 
> Did cppcheck have anything else to say?
> 
> Thanks!
> sage

-- 
GMX DSL: Internet, Telefon und Entertainment für nur 19,99 EUR/mtl.!
http://portal.gmx.net/de/go/dsl02

-- 
Sicherer, schneller und einfacher. Die aktuellen Internet-Browser -
jetzt kostenlos herunterladen! http://portal.gmx.net/de/go/chbrowser