libgcj/9078: libffi: problems with uint8 on powerpc

[Matthias Klose]

The following reply was made to PR libgcj/9078; it has been noted by GNATS.

From: Matthias Klose <doko@cs.tu-berlin.de>
To: Jeff Sturm <jsturm@one-point.com>
Cc: 173074-done@bugs.debian.org, gcc-gnats@gcc.gnu.org, <gcc-prs@gcc.gnu.org>,
        <gcc-bugs@gcc.gnu.org>, <java-prs@gcc.gnu.org>,
        David Paul BELANGER <dbelan2@CS.McGill.CA>,
        "Prof. Etienne M. Gagnon" <etienne.gagnon@uqam.ca>
Subject: Re: libgcj/9078: libffi: problems with uint8 on powerpc
Date: Sun, 16 Mar 2003 18:02:34 +0100

 Ok, closing the Debian report. Leave the gnats entry to Jeff.
 
 Prof. Etienne M. Gagnon writes:
 > On Sun, Mar 16, 2003 at 05:15:47PM +0100, Matthias Klose wrote:
 > > hmm, you didn't open this report. Sure I can close it?
 > 
 > Yes, I am sure.  The bug was opened by of the developers of the
 > SableVM project, and he forwarded to me the information in the Debian
 > BTS and in the upstream BTS.  Based on that information, I went back
 > into the SableVM source code and fixed the bug.  It was definitely not
 > a libffi bug (yet one has to very carfully read the libffi
 > documentation to notice the inconsistency in the treatment of function
 > arguments and return values).
 
 
 Jeff Sturm writes:
 > On Sat, 28 Dec 2002, Matthias Klose wrote:
 > > ffi_type_uint8 and other arguments shorter than one word are
 > > not passed in correctly to the function called by ffi_call.
 > 
 > For an ffi_type_uint8, ffi_call expects the corresponding value
 > pointer to be a (unsigned char *).  So this is correct usage:
 > 
 >     int b = 24;
 >     args[0] = &ffi_type_uint;
 >     values[0] = &b;
 > 
 > Also correct would be:
 > 
 >     unsigned char b = 24;
 >     args[0] = &ffi_type_uint8;
 >     values[0] = &b;
 > 
 > > Also, returns values are not passed correctly.  For example, instead of
 > > finding the returned byte value where the pointer points to, it is found
 > > at an offset of 3 from the pointer.
 > 
 > Return values are handled a little differently than arguments.
 > libffi/README says:
 > 
 >     RVALUE is a pointer to a chunk of memory that is to hold the
 >         result of the function call. Currently, it must be
 >         at least one word in size (except for the n32 version
 >         under Irix 6.x, which must be a pointer to an 8 byte
 >         aligned value (a long long). It must also be at least
 >         word aligned (depending on the return type, and the
 >         system's alignment requirements). If RTYPE is
 >         &ffi_type_void, this is ignored. If RVALUE is NULL,
 >         the return value is discarded.
 > 
 > So this cannot work:
 > 
 >     unsigned char result;
 >     if (ffi_prep_cif(&cif, FFI_DEFAULT_ABI, 1,
 >   		   &ffi_type_uint8, args) != FFI_OK) {
 > 
 > You could use "unsigned int result" on a 32-bit target, or use ffi_arg
 > which is typedef'ed to work correctly on 32 or 64-bit targets:
 > 
 >     ffi_arg result;
 >     if (ffi_prep_cif(&cif, FFI_DEFAULT_ABI, 1,
 >                    &ffi_type_uint8, args) != FFI_OK) {
 > 
 > With those changes your example should be portable to any target supported
 > by libffi.
 > 
 > Jeff
 > 
 > 
 > -- 
 > To UNSUBSCRIBE, email to debian-gcc-request@lists.debian.org
 > with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org