On Sat, 2010-07-03 at 12:35 +0200, Bernd Roesch wrote:

> So the 
> 
> * Failed writing body (0 != 218) 
> 
> messages are ok and do not make introduce a speed penalty ?

Yes, they are fine. They result from us aborting the fetch without
processing any data when we detect that we've already got a suitable
object in the cache.

> > Libcss claims it is not receiving all the input data. NetSurf is passing
> > it all the data, and using the libcss API correctly. Therefore, it's
> > likely an issue with the input stream.
> > 
> > What version of libparserutils are you using?
> 
> it was the newest version (lib not updatet since end of 2009).seem compiled 
> with GCC 3.4.0
> 
> I checkout files now and compile it with GCC 4.5.0 -O2 and the error message 
> is gone.

Right.

> I think in the makefiles you should add a warning notice, that compiler > 3.4 
> should use.
> I see that more often that GCC 3.4.0 create buggy code in netsurf so i 
> suggest always compile
> with GCC 4.3 or higher netsurf.

Or not. NetSurf has been built with many versions of GCC over the years,
and almost none of them result in broken binaries. At least one of the
developers is still using GCC 3.4 to compile NetSurf and associated
libraries, and has not experienced the problems you are seeing. Thus, it
sounds to me as if any GCC 3.4-specific problems you are encountering
are specific to the platform you are targetting or the CPU architecture
that platform uses.

> Have somebody compile whole netsurf with GCC 3.4 and it work ?

Yes, we did so for years without issue (and some still do, as I said
above)
 
> > Have you enabled the iconv input filter in libparserutils?
> 
> I dont know, i change nothing, i compile only with make and i get this output.

Right, so the answer to my question is "no".


J.


Reply via email to