On Sun, 2008-05-11 at 17:01 +0200, David Eriksson wrote: > > On Sun, 2008-05-11 at 10:55 +0100, Mark Ellis wrote: > > On Fri, 2008-05-09 at 22:40 +0200, David Eriksson wrote: > > > > > > On Tue, 2008-05-06 at 21:50 +0100, Mark Ellis wrote: > > > > Hi All > > > > > > > > General inquiry, does anyone know why CeReadRecordProps and > > > > CeWriteRecordProps for rapi1 are inside an #if SIZEOF_VOID_P == 4 ? > > > > > > > > After a brief browse I can't see anything obvious that would break. I > > > > also can't see this defined anywhere. > > > > > > SIZEOF_VOID_P is defined in rapi_config.h (generated by the configure > > > script, i.e. AC_CHECK_SIZEOF(void *) in configure.ac.in) > > > > > > If this #define is of my making, I think that it is because it is very > > > important that CEVALUNION.lpwstr is read and written as a 32-bit value > > > and same for CEVALUNION.blob.lpb > > > > > > > Thanks David, that should give me something to think about for a while. > > For the sake of clarity: this would of course only be an issue on > platforms where a pointer is 64-bit. > >
I was expecting the problem here to be that interesting bit of pointer arithmetic, but it seems not. On 64 bit, CEVALUNION is 16 bytes long, because CEBLOB is 16 bytes. CEVALUNION over the wire is 8 bytes. I guess the problem is because CEBLOB is a struct of 2 members, each member gets allocated on 64 bit boundaries, hence a 16 byte struct, even though they don't need this much space. Compiler quirks aren't my strength, do you know how to force this to the correct size ? Mark ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ SynCE-Devel mailing list SynCE-Devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/synce-devel
