AW: [Gnash-commit] gnash ChangeLog testsuite/swfdec/PASSING testsu...

[Benjamin Wolsey]

> > If everyone is happy with boost::format - we've encountered most of its 
> problems by now, I 
> > hope - I would start replacing SIZET_FMT and not using c_str(); I'd rather 
> > not 
> do that until
> > there's some sort of agreement, though.
> 
> I was thinking the same thing..
> Maybe let's wait for Hong to find out what's the problem with
> stringstream on ARM. 
> And I guess  you did already take the easy compile-time switch to run
> performance comparisons, right ?
> 
> --strk;
>

The switch is in log.h (#define USE_BOOST_FORMAT_TEMPLATES 1) and has been 
enabled by default since the 0.8.2 release (in case anyone else wants to run 
some tests - my tests are documented in 
http://wiki.gnashdev.org/wiki/index.php/Logging). The advantages and 
disadvantages of using boost::format are also listed there.

Boost::format is clearly much slower when extremely large amounts of logging 
are done, for instance running with -va (but it allows much longer messages); 
under normal circumstances the difference seems negligible.

I don't know if this also applies to other (e.g. embedded) devices - if there 
is a noticable performance difference under ordinary circumstances I'd like to 
know about it.

It looks like Hong Yu's problem is in stringstream rather than boost::format, 
but there's no problem with waiting for that to be sorted first.

bwy.





      Machen Sie Yahoo! zu Ihrer Startseite. Los geht's: 
http://de.yahoo.com/set