Dan Sugalski wrote:
> At 01:07 PM 3/1/00 -0800, [EMAIL PROTECTED] wrote:
> >Dan Sugalski wrote:
> > > On the other hand, other folks on other platforms have reported problems
> > > with the version variable, and that's where this could be dying.
> >
> >Are you referring to $] ?
>
> Yup. It looks OK in .670 and .660 here, but it apparently goes odd other
> places.
Perhaps it's a api_version || api_revision || api_subversion ||
api_versionstring || pm_apiversion || xs_apiversion problem?
> > > Does it still go in .670? (Yeah, I know--in what free time? :)
> >
> >Here is the tail end of a @configure "-des" build with DEC C V5.3-006
> >on OpenVMS Alpha V6.2-1H3:
> >
> >%LINK-W-NUDFSYMS, 1 undefined symbol:
> >%LINK-I-UDFSYM, H_ERRNO
> >
> >OK in this dirty build tree I simply remove the extern from h_errno in
> >pp_sys.c
> >and then see:
>
> I bet it is h_errno causing the problem. Probably some wacko pointer
> someplace that just hoses everything. (I hope)
>
> I'm trying to get going on the book, so I'll look at some way around this
> tomorrow, I think.
The problem could be with h_errno for all I know. I was able to build 5.005_02
by simply removing the extern. I think that I also tried messing with undefing
HOST_NOT_FOUND as you had suggested, but I might have done that in
vms/sockadapt.h in my previous test.
If I do this unconditional butchery to pp_sys.c:
--- DKA200:[PERL-5_5_670]PP_SYS.C;1 Sun Feb 27 18:07:32 2000
+++ pp_sys.c Thu Mar 2 11:20:57 2000
@@ -77,6 +77,9 @@
compiling multithreaded and singlethreaded ($ccflags et al).
HOST_NOT_FOUND is typically defined in <netdb.h>.
*/
+#if defined(HOST_NOT_FOUND)
+# undef HOST_NOT_FOUND
+#endif
#if defined(HOST_NOT_FOUND) && !defined(h_errno) && !defined(__CYGWIN__)
extern int h_errno;
#endif
End of diff (NOT a patch).
Then I obtain, after a @configure "-des" setup with 5.5.670:
Copy/Log/NoConfirm [.VMS.EXT]XSSYMSET.PM [.LIB.EXTUTILS]XSSYMSET.PM
%COPY-S-COPIED, DKA200:[PERL-5_5_670.VMS.EXT]XSSYMSET.PM;1 copied to
DKA200:[PERL-5_5_670.LIB.EXTUTILS]XSSYMSET.PM;1 (15 blocks)
MCR Sys$Disk:[]miniperl.exe "-I[.lib]" "-I[.ext.re]" [.lib.extutils]xsubpp
-noprototypes [.EXT.DYNALOADER]DL_VMS.XS >[.EXT.DYNALOADER]DL_VMS.C
%SYSTEM-F-ACCVIO, access violation, reason mask=00, virtual address=00000007,
PC=80566578, PS=0000001B
%TRACE-F-TRACEBACK, symbolic stack dump follows
image module routine line rel PC abs PC
0 0000000000000000 FFFFFFFF80566578
MINIPERL SV Perl_sv_vcatpvfn 52472 000000000000E608 00000000000CAFD8
MINIPERL SV Perl_sv_vsetpvfn 52204 000000000000DB8C 00000000000CA55C
MINIPERL SV Perl_sv_vsetpvf 52093 000000000000D840 00000000000CA210
MINIPERL SV Perl_sv_setpvf 52087 000000000000D7D0 00000000000CA1A0
MINIPERL PERL S_incpush 52610 0000000000007670 00000000000578B0
MINIPERL PERL S_parse_body 47588 0000000000001D48 0000000000051F88
MINIPERL PERL perl_parse 47453 0000000000001700 0000000000051940
MINIPERL MINIPERLMAIN main 46580 0000000000000164 0000000000050164
MINIPERL MINIPERLMAIN __main 0 0000000000000070 0000000000050070
0 FFFFFFFF85122170 FFFFFFFF85122170
%MMS-F-ABORT, For target [.EXT.DYNALOADER]DL_VMS.C, CLI returned abort status:
%X1000000C.
Perhaps we ought to come up with an acceptable patch to pp_sys.c (dependent on
__VMS_VERSION if necessary) then recommend a VMS debug build for folks with
DEC C V5.3-006 on OpenVMS Alpha V6.2-1H3 as well as with Charles Bailey's
version of DECC. How does that grab folks? (Note that I am not too dependent
on this compiler as I can upgrade on this particular box).
Peter Prymmer
