From:	MERC::"uunet!WKUVX1.BITNET!MacroMan"  1-DEC-1992 19:17:50.24
To:	gjc@mitech.com, <macro32%wkuvx1.bitnet%cunyvm.cuny.edu@TGV.COM>,
	<alpha-ids%wkuvx1.bitnet%cunyvm.cuny.edu@TGV.COM>
CC:	
Subj:	Re: MEMBER_ALIGHMENT difficulties on ALPHA?

In article <3865@mitech.com> you write:
#
#Since there is no vmsnet.alpha group yet, I figured the best audience to
#post this one to is vmsnet.internals.
#

Well there is a mail list: alpha-ids@wkuvx1.bitnet

#
#Boy am I having problems with something like MEMBER_ALIGNMENT issues on Alpha,
#and perhaps it has something to do with having <stat.h> in the list of
#include files.
#

<stat.h> should not cause any difficulties aside from the fact that the stat
structure needs to be surround with the following

        #pragma member_alignment save   /* saves the current alignment state */
        #pragma member_noalignment      /* turns off alignment */
          [ stat structure]
        #pragma member_alignment restore /* restores the alignment state */
#
#Two files, OBJ_UTIL.C and PARSE.C both include the definition of
#struct obj. But the structure is treated differently and looks differently
#in the debugger inside each file.
#
#   904:  op = (struct object *)palloc(sizeof(struct object));
#    DBG> e op->o_u.KB
#    OBJ_UTIL\new_object\op->o_u.KB
#        cur_time:   -100000000000.0000
#        cur_value:  0.0000000000000000
#
#But if you single step out of the procedure new_object into its caller
#in a different file:
#
#    stepped to PARSE\parse\parse_1\%LINE 7458+10
#      7458:         op = new_object(O_KB);
#    DBG> s
#    stepped to PARSE\parse\parse_1\%LINE 7459
#    DBG> e op->o_u.KB
#    PARSE\parse\op->o_u.KB
#        cur_time:   0.0000000000000000
#        cur_value:  -100000000000.0000
#
#
#It is as if the order of the slots has been reversed. So, probably this
#has nothing to do with MEMBER_ALIGNMENT, and is some other problem.
#

Well it might, depends on what the previous 4 bytes in memory contain before
-100000000000.0000.  It would be very helpful to see the definition of the
structure obj, also the see the function definition of new_object().  I am
unable to tell what you are passing to new_object().  If it is a structure,
then this is a known problem with the DEC C compiler.  Also, to guarantee
that both modules are including the same definition.

#
#I found that keeping <stat.h> out of the list of include files fixed
#the problem for some files.
#

This should only affect it iff you are using the stat structure as part of
the obj structure.

--
Patrick L. Mahan

--- TGV Window Washer ------------------------------- Mahan@TGV.COM ---------

Waking a person unnecessarily should not be considered  - Lazarus Long
a capital crime.  For a first offense, that is            From the Notebooks of
                                                          Lazarus Long