> I also finally got a chance to look at this code. I've included my
> comments below.
My primary concern is nto with Pythin bits but with ./modules/arch.c
in which the source can not be compiled on Solaris 8 unless I do
some fairly blunt force things with it :
#include <sys/systeminfo.h>
/************************************************************************/
/* This is an outright blatant hack by Dennis Clarke. A blunt force */
/* trauma method to simply ensure that the values in sys/systeminfo.h */
/* are actually defined. These values are missing in Solaris 8 */
/* */
/* See sys/systeminfo.h in Solaris Nevada snv_70b for further info */
/************************************************************************/
/*
* These values have to be registered
* with Unix International can't be corrected now. The status of a command
* as published or unpublished does not alter the algorithm.
*/
#define SI_ARCHITECTURE_32 516 /* basic 32-bit SI_ARCHITECTURE */
#define SI_ARCHITECTURE_64 517 /* basic 64-bit SI_ARCHITECTURE */
#define SI_ARCHITECTURE_K 518 /* kernel SI_ARCHITECTURE equivalent */
#define SI_ARCHITECTURE_NATIVE 519 /* SI_ARCHITECTURE of the caller */
These are needed in order to satisfy these things further down :
/*
* Return a list of strings constituting the architecture tags for the invoking
* system.
*/
/*ARGSUSED*/
static PyObject *
arch_isainfo(PyObject *self, PyObject *args)
{
char *buf1;
char *buf2;
char *buf = NULL;
PyObject *robj;
buf1 = get_sysinfo(SI_ARCHITECTURE_64);
buf2 = get_sysinfo(SI_ARCHITECTURE_32);
still further down I see things like this :
/*
* Return the platform tag ("i86pc") for the invoking system.
*/
/*ARGSUSED*/
static PyObject *
arch_platform(PyObject *self, PyObject *args)
{
char *buf = NULL;
PyObject *robj;
buf = get_sysinfo(SI_PLATFORM);
if (buf == NULL)
return (NULL);
robj = Py_BuildValue("s", buf);
free(buf);
return (robj);
}
... where I try to figure out the thinking involved and I wonder if
"i86pc" is the only value that will be returned or what. I am not sure
and wading through source isn't helping me in my efforts to back port
this to Solaris 8. Going to Solaris 10 loks to be no big deal but further
back is tougher. Certainly when we consider the issues with the libmd
libs in S10u4 that simply do not exist in S10GA or anything previous.
Everything that needs the crypto services satisfied in these libs *need*
to be solved with OpenSSL calls.
But .. I'm just flailing away here. Not making much progress.
Am I on target ?
Dennis
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
