10 Dec
2019
10 Dec
'19
4:25 p.m.
I'm trying to convert some C code[1] so it'll compile on TOPS20 with KCC.
KCC is mostly ANSI compliant, but it needs to use the TOPS20 linker, which
has a limit of six case-insentive characters. Adam Thornton wrote a Perl
script[2] that successfully does this for Frotz 2.32. The Frotz codebase
has evolved past what was done there and so 2.50 no longer works with
Adam's script. So I've been expanding that script into something of my
own, which I call "snavig"[3]. It seems to be gradually working more and
more, but I fear the problem is starting to rapidly diverge because it
still doesn't yield compilable code even on Unix. Does anyone here have
any knowledge of existing tools or techniques to do what I'm trying to do?
This is part of a project to get Infocom and other Z-machine games running
once again on PDP10 mainframes, either real or emulated. First up is to
get the bare minimum of a current Z-machine emulator running on TOPS20.
Then we can work on screen-handling, a disk pager[4], and porting to other
PDP10 operating systems. I'm hoping that this will lead to fun exhibits
wherever PDP10s are displayed in museum or faire settings.
[1] https://gitlab.com/DavidGriffith/frotz
[2] https://github.com/athornton/gnusto-frotz-tops20
[3] Change an objects shape.
[4] Infocom's Z-machine emulators paged zcode from disk, but Frotz simply
sucks the whole zcode file into memory.
--
David Griffith
dave at 661.org
A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
10 Dec
10 Dec
4:48 p.m.
any chance you can post-process the .obj files?
Warner
On Tue, Dec 10, 2019 at 5:26 PM David Griffith via cctalk <
cctalk at classiccmp.org> wrote:
I'm trying to convert some C code[1] so it'll compile on TOPS20 with KCC.
KCC is mostly ANSI compliant, but it needs to use the TOPS20 linker, which
has a limit of six case-insentive characters. Adam Thornton wrote a Perl
script[2] that successfully does this for Frotz 2.32. The Frotz codebase
has evolved past what was done there and so 2.50 no longer works with
Adam's script. So I've been expanding that script into something of my
own, which I call "snavig"[3]. It seems to be gradually working more and
more, but I fear the problem is starting to rapidly diverge because it
still doesn't yield compilable code even on Unix. Does anyone here have
any knowledge of existing tools or techniques to do what I'm trying to do?
This is part of a project to get Infocom and other Z-machine games running
once again on PDP10 mainframes, either real or emulated. First up is to
get the bare minimum of a current Z-machine emulator running on TOPS20.
Then we can work on screen-handling, a disk pager[4], and porting to other
PDP10 operating systems. I'm hoping that this will lead to fun exhibits
wherever PDP10s are displayed in museum or faire settings.
[1] https://gitlab.com/DavidGriffith/frotz
[2] https://github.com/athornton/gnusto-frotz-tops20
[3] Change an objects shape.
[4] Infocom's Z-machine emulators paged zcode from disk, but Frotz simply
sucks the whole zcode file into memory.
--
David Griffith
dave at 661.org
A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
4:53 p.m.
It was thus said that the Great David Griffith via cctalk once stated:
I'm trying to convert some C code[1] so it'll compile on TOPS20 with KCC.
KCC is mostly ANSI compliant, but it needs to use the TOPS20 linker, which
has a limit of six case-insentive characters. Adam Thornton wrote a Perl
script[2] that successfully does this for Frotz 2.32. The Frotz codebase
has evolved past what was done there and so 2.50 no longer works with
Adam's script. So I've been expanding that script into something of my
own, which I call "snavig"[3]. It seems to be gradually working more and
more, but I fear the problem is starting to rapidly diverge because it
still doesn't yield compilable code even on Unix. Does anyone here have
any knowledge of existing tools or techniques to do what I'm trying to do?
If you are doing this on Linux, an approach is to compile the code there,
then run 'nm' over the object files, and it will output something like:
[spc]lucy:~/source/boston/src>nm main.o
000000ef t CgiMethod
U CgiNew
00000000 r __PRETTY_FUNCTION__.21
U __assert_fail
U crashreport_core
U crashreport_with
U gd
U gf_debug
00000000 T main
U main_cgi_get
U main_cgi_head
U main_cgi_post
U main_cli
The last column are identifiers; the second column is the type of
identifier, and the first column is the value. What you want to look for
are types 'T' (externally visible function), 'C' (externally visible
constant data) and 'D' (externally visible data). It is these identifiers
that will need to be six unique characters long.
Something like:
[spc]lucy:~/source/boston/src>nm globals.o | grep ' [CDT] '
00000041 T GlobalsInit
00000004 C c_adtag
00000004 C c_class
00000004 D c_conversion
00000004 C c_days
00000004 C c_tzmin
00000000 D c_updatetype
00000004 C c_webdir
00000008 D cf_emailupdate
00000004 C g_L
00000004 C g_blog
00000004 C g_templates
00000020 D gd
00000d09 T set_c_conversion
00000beb T set_c_updatetype
00000dbd T set_c_url
00000cab T set_cf_emailupdate
(but over all object files). I would then generate unique six character
long identifiers for each of these, and slap the output into a header file
like:
#define GlobalsInit id0001
#define c_adtag id0002
#define c_class id0003
#define c_conversion id0004
and then include this file for every compilation unit. I think that would
be the easiest thing to do.
-spc
4:54 p.m.
On Tue, Dec 10, 2019 at 5:53 PM Sean Conner via cctalk <
cctalk at classiccmp.org> wrote:
It was thus said that the Great David Griffith via
cctalk once stated:
You'd need to exclude libc symbols, though.
In a.out times you could just change the names from old to new and adjust
the string table offset to do this :)
Warner
I'm trying to convert some C code[1] so it'll compile on TOPS20 with
KCC.
KCC is mostly ANSI compliant, but it needs to use
the TOPS20 linker,
which
has a limit of six case-insentive characters.
Adam Thornton wrote a
Perl
script[2] that successfully does this for Frotz
2.32. The Frotz
codebase
has evolved past what was done there and so 2.50
no longer works with
Adam's script. So I've been expanding that script into something of my
own, which I call "snavig"[3]. It seems to be gradually working more
and
more, but I fear the problem is starting to
rapidly diverge because it
still doesn't yield compilable code even on Unix. Does anyone here have
any knowledge of existing tools or techniques to do what I'm trying to
do?
If you are doing this on Linux, an approach is to compile the code there,
then run 'nm' over the object files, and it will output something like:
[spc]lucy:~/source/boston/src>nm main.o
000000ef t CgiMethod
U CgiNew
00000000 r __PRETTY_FUNCTION__.21
U __assert_fail
U crashreport_core
U crashreport_with
U gd
U gf_debug
00000000 T main
U main_cgi_get
U main_cgi_head
U main_cgi_post
U main_cli
The last column are identifiers; the second column is the type of
identifier, and the first column is the value. What you want to look for
are types 'T' (externally visible function), 'C' (externally visible
constant data) and 'D' (externally visible data). It is these identifiers
that will need to be six unique characters long.
Something like:
[spc]lucy:~/source/boston/src>nm globals.o | grep ' [CDT] '
00000041 T GlobalsInit
00000004 C c_adtag
00000004 C c_class
00000004 D c_conversion
00000004 C c_days
00000004 C c_tzmin
00000000 D c_updatetype
00000004 C c_webdir
00000008 D cf_emailupdate
00000004 C g_L
00000004 C g_blog
00000004 C g_templates
00000020 D gd
00000d09 T set_c_conversion
00000beb T set_c_updatetype
00000dbd T set_c_url
00000cab T set_cf_emailupdate
(but over all object files). I would then generate unique six character
long identifiers for each of these, and slap the output into a header file
like:
#define GlobalsInit id0001
#define c_adtag id0002
#define c_class id0003
#define c_conversion id0004
and then include this file for every compilation unit. I think that would
be the easiest thing to do.
11 Dec
11 Dec
9:07 a.m.
On Wed, 2019-12-11 at 00:25 +0000, David Griffith via cctalk wrote:
I'm trying to convert some C code[1] so it'll
compile on TOPS20 with KCC.
KCC is mostly ANSI compliant, but it needs to use the TOPS20 linker, which
has a limit of six case-insentive characters. [...] Does anyone here have
any knowledge of existing tools or techniques to do what I'm trying to do?
Is "objcopy --redefine-syms" any help? Compile the code as-is to
produce object files, use nm or objdump to find all of the global
symbols, generate unique six-character names for them, and then use
objcopy to create new object files with the new names.
Or have I completely missed the point? I'm not familiar with KCC, does
it produce object modules in a format objcopy doesn't support?
I know someone who was working on gcc support for the PDP-10, I wonder
if he's still doing that or has given up....
11:42 a.m.
I think the challenge will be does binutils (where nm, objcopy and objdump live) support
for the object file format used by TOPS20.
I haven?t looked at the TOPS20 object file format but it seems like the best approach
would be to have the C compiler generate symbols as it normally would and write a utility
to ?fixup? the too long symbols rather than munging the source (which is basically what
you?re proposing using the stuff from binutils?just a bit more work.
TTFN - Guy
On Dec 11, 2019, at 9:07 AM, Guy N. via cctalk
<cctalk at classiccmp.org> wrote:
On Wed, 2019-12-11 at 00:25 +0000, David Griffith via cctalk wrote:
I'm trying to convert some C code[1] so
it'll compile on TOPS20 with KCC.
KCC is mostly ANSI compliant, but it needs to use the TOPS20 linker, which
has a limit of six case-insentive characters. [...] Does anyone here have
any knowledge of existing tools or techniques to do what I'm trying to do?
Is "objcopy --redefine-syms" any help? Compile the code as-is to
produce object files, use nm or objdump to find all of the global
symbols, generate unique six-character names for them, and then use
objcopy to create new object files with the new names.
Or have I completely missed the point? I'm not familiar with KCC, does
it produce object modules in a format objcopy doesn't support?
I know someone who was working on gcc support for the PDP-10, I wonder
if he's still doing that or has given up....
11:37 a.m.
KCC is mostly ANSI compliant, but it needs to use the
TOPS20 linker, which
has a limit of six case-insentive characters.
LINK has support for long (up to 72 character) symbols, and it appears
FORTRAN v11 can generate them, but the MACRO assembler may never have
gotten support;
http://pdp-10.trailing-edge.com/bb-r775d-bm_tops20_ks_upd_4/01/sources/lnkn…
SUBTTL BLOCK TYPE 1002 - LONG SYMBOL ENTRY
; ----------------
; ! 1002 ! COUNT !
; ----------------
; ! SYMBOL !
; ----------------
; ! MORE SYMBOL !
; ----------------
in lnkpar.mac:
ND MAXSYM,^D72 ;[2210] Max number of characters in a symbol
ND SYMSIZ,<<MAXSYM+5>/6> ;[2210] Number of words of storage for long symbols
1607
days inactive
1607
days old
6 comments
6 participants
participants (6)
-
dave＠661.org -
decguy＠songdog.eskimo.com -
ggs＠shiresoft.com -
imp＠bsdimp.com -
phil＠ultimate.com -
spc＠conman.org