On Tue, Oct 02, 2007 at 12:06:17AM -0700, Steve Langasek wrote:
> On Tue, Oct 02, 2007 at 08:32:40AM +0200, Michael Koch wrote:
> > On Tue, Oct 02, 2007 at 12:21:31AM +0100, Ben Hutchings wrote:
> > > diff -u <(c++filt <libskstream-0.3-4_i386 | sort) <(c++filt
> > > <libskstream-0.3-4_amd64 | sort)
> > > --- /dev/fd/63 2007-10-02 00:19:14.445928644 +0100
> > > +++ /dev/fd/62 2007-10-02 00:19:14.445928644 +0100
> > > @@ -102,7 +102,7 @@
> > > raw_socket_stream::raw_socket_stream(FreeSockets::IP_Protocol)@Base
> > > 0.3.6-2
> > > raw_socket_stream::setBroadcast(bool)@Base 0.3.6-2
> > > raw_socket_stream::setProtocol(FreeSockets::IP_Protocol)@Base 0.3.6-2
> > > - socketbuf::setbuf(char*, int)@Base 0.3.6-2
> > > + socketbuf::setbuf(char*, long)@Base 0.3.6-2
> > > socketbuf::setSocket(int)@Base 0.3.6-2
> > > socketbuf::~socketbuf()@Base 0.3.6-2
> > > socketbuf::~socketbuf()@Base 0.3.6-2
>
> > That was exactly what I was referring to. Sorry for being unclear.
>
> Oh, yes, ok.
>
> > Is it a common problem that APIs use some 32-bit data types on 32-bit
> > archs/userland and 64-bit data types on 64-bit archs?
>
> Well, on a 32-bit arch, a long is a 32-bit type; on a 64-bit arch, a long is
> a 64-bit type. And AIUI, on 32-bit archs a "long" is called an "int" in C++
> symbol-mangling.
In may case the culprit type is std::streamsize which is defined by
ptrdiff_t and this is arch specific. That is way the setBuf method has
different definition:
32-bit archs:
[EMAIL PROTECTED] 0.3.6-2
64-bit archs:
[EMAIL PROTECTED] 0.3.6-2
Note the 'i' vs 'l'.
Cheers,
Michael
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
