axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Davanum Srinivas <dava...@gmail.com>
Subject Re: Fatal bug in UnsignedLong
Date Tue, 04 May 2004 11:31:27 GMT
Yes, please open a JIRA bug with your patch asap
(http://nagoya.apache.org/jira).

thanks,
dims

On Tue,  4 May 2004 12:54:26 +0200, tommy@alander.se <tommy@alander.se> wrote:
> 
> 
> Quoting Steve Loughran <steve_l@iseran.com>:
> 
> > tommy@alander.se wrote:
> > > Hi,
> > >
> > > There is a fatal bug in the UnsignedLong implementation. It currently uses
> >
> > > java.lang.Double as its storage. This doesn't work.
> > >
> > > For example if you recieve the value 35920361897405533 this value will
> > > be "transformed" to the incorrect value of 35920361897405532.
> > >
> > > I have made an alternative impl using BigInteger. Should I send a patch?
> > >
> > > /Tommy
> > >
> > >
> > Yes, with tests. But I wouldnt expect it to be used until Axis1.3
> > because it is quite a major change in API use in code (maybe we need a
> > prewarning in Axis1.2 that it will change). JAX-RPC doesnt cover ulong
> > to java type mapping; as far as Sun are concerned there are no unsigned
> > longs.
> 
> The internal API remains the same. Only the internal storage is changed to
> use BigInteger instead of Double.
> 
> IMHO this needs to be changed before 1.2 since it corrupts the data.
> 
> /Tommy
> 
>

Mime
View raw message