I'm trying to resurrect a VAX MACRO program I wrote in 1988: VTIL, a VAX Threaded Interpretive Language. It's a Forth and STOIC inspired language that uses subroutine threading and it was my senior project in college. I don't have the media it was saved on, or the devices to read those media, but I do have a printout of the VAX MACRO source. I don't have the command procedures I would have used to compile it.

I've got a VAX/VMS 5.5-2 system that I maintain these days (it's emulated) that is a descendant of one of the VAXes I wrote this code on originally. While I've been programming regularly on VMS over the years, most of it has been in VAX BASIC or GNU C. I haven't written any serious MACRO code, probably since ... 1988.

I've typed the MACRO source all in, and got rid of all the obvious errors resulting from the process, and it assembles without errors. When I went to link it, I found out that it was missing two external symbols, which from context I figured out were for items from a VAX Message file that was missing (it wasn't on the printout, alas). I constructed a new .MSG file that ought to work, but when I tried to link them I got the following link error:

$ LINK/MAP/FULL/EXE=VTIL.EXE VTIL.OBJ,VTIL_MSG.OBJ
%LINK-E-OUTSIMG, attempted store location %X00001FFC is outside image binary (%X00002000 to %X02001E00)
in psect VTILCODE module VTIL file MPL$DATA3[MPL.TKB.VTIL]VTIL.OBJ;1
-LINK-E-NOIMGFIL, image file not created
%MMK-F-ERRUPD, error status %X106482B2 occurred when updating target VTIL.EXE

I was hoping someone could give me some guidance on how to figure out what is happening here.

The code is available on GitHub, if anyone wants to read it:
https://github.com/tkurtbond/VTIL

Writing a Forth-inspired language on a VAX using subroutine threading is definitely playing the Forth implementation game in easy mode. I was pleased to see that I could understand the code, after I looked up a few instructions. It is pretty straightforward.

I'd say that if I get it working again it would need some additions to make it generally useful, mainly general file i/o. Right now it can load VTIL source from text files and open a block file and use Forth-inspired functions for reading and writing blocks. It was originally designed to use LIB$GET_INPUT and LIB$PUT_OUTPUT for terminal i/o, and I was thinking it would be interesting to see if I could modify it to have a more Forth-like approach to terminal input and output, where it is done a character at a time.

Since you have LINK/MAP, you could post the link map.

It looks like an external symbol problem of some kind, where the linker is trying to modify an address, but the wrong place.

Interesting, it is just before what seems to be the beginning.

On Saturday, August 12, 2023 at 2:30:01 AM UTC+1, gah4 wrote:
> Since you have LINK/MAP, you could post the link map.
>
> It looks like an external symbol problem of some kind, where the linker is trying to modify an address, but the wrong place.
>
> Interesting, it is just before what seems to be the beginning.

There's something weird in the map output anyway.
The cluster listings don't make sense and there seem to be a lot of rows for the default cluster.
VTILCODE is also being listed as 4Gb in size roughly.
I noticed that the beginning of VTILCODE, the initial macro expansion has an offset of -4 and 00001FFC is suspiciously 4 away from 00002000
Possibly barking up the wrong tree but I'll have a proper look later.

K

On Saturday, August 12, 2023 at 6:46:11 AM UTC-4, cao[...] wrote:
> On Saturday, August 12, 2023 at 2:30:01 AM UTC+1, gah4 wrote:
> > Since you have LINK/MAP, you could post the link map.
> >
> > It looks like an external symbol problem of some kind, where the linker is trying to modify an address, but the wrong place.
> >
> > Interesting, it is just before what seems to be the beginning.
> There's something weird in the map output anyway.
> The cluster listings don't make sense and there seem to be a lot of rows for the default cluster.
> VTILCODE is also being listed as 4Gb in size roughly.
> I noticed that the beginning of VTILCODE, the initial macro expansion has an offset of -4 and 00001FFC is suspiciously 4 away from 00002000
> Possibly barking up the wrong tree but I'll have a proper look later.
>
> K

As cao noticed, the MAP file from the most recent link *is* in the https://github.com/tkurtbond/VTIL repository and the direct link is https://github.com/tkurtbond/VTIL/blob/main/vtil.map

Sorry, I should have mentioned that.

On 8/12/2023 11:11 AM, T. Kurt Bond wrote:
> On Saturday, August 12, 2023 at 6:46:11 AM UTC-4, cao[...] wrote:
>> On Saturday, August 12, 2023 at 2:30:01 AM UTC+1, gah4 wrote:
>>> Since you have LINK/MAP, you could post the link map.
>>>
>>> It looks like an external symbol problem of some kind, where the linker is trying to modify an address, but the wrong place.
>>>
>>> Interesting, it is just before what seems to be the beginning.
>> There's something weird in the map output anyway.
>> The cluster listings don't make sense and there seem to be a lot of rows for the default cluster.
>> VTILCODE is also being listed as 4Gb in size roughly.
>> I noticed that the beginning of VTILCODE, the initial macro expansion has an offset of -4 and 00001FFC is suspiciously 4 away from 00002000
>> Possibly barking up the wrong tree but I'll have a proper look later.
>>
>> K
>
> As cao noticed, the MAP file from the most recent link *is* in the https://github.com/tkurtbond/VTIL repository and the direct link is https://github.com/tkurtbond/VTIL/blob/main/vtil.map
>
> Sorry, I should have mentioned that.

I'm impressed that you somehow crafted a PSECT with a negative length!

Psect Name Module Name Base End Length
---------- ----------- ---- --- ------

VTILCODE 00002000 00001FFB FFFFFFFC ( 4294967292.)
VTIL 00002000 00001FFB FFFFFFFC ( 4294967292.)

We'll need John Reagan to weigh in here.

--

--- Rob

On Saturday, August 12, 2023 at 1:32:27 PM UTC-4, Robert A. Brooks wrote:

> > As cao noticed, the MAP file from the most recent link *is* in the https://github.com/tkurtbond/VTIL repository and the direct link is https://github.com/tkurtbond/VTIL/blob/main/vtil.map
> >
> > Sorry, I should have mentioned that.
> I'm impressed that you somehow crafted a PSECT with a negative length!
>
>
> Psect Name Module Name Base End Length
> ---------- ----------- ---- --- ------
>
> VTILCODE 00002000 00001FFB FFFFFFFC ( 4294967292.)
> VTIL 00002000 00001FFB FFFFFFFC ( 4294967292.)
>
>
> We'll need John Reagan to weigh in here.

Well, that doesn't look good. I've updated that repository with the .MAP and VTIL.LIS with /SHOW=EXPANSION from a compilation done today. It shouldn't be too different.

On Saturday, August 12, 2023 at 3:39:12 PM UTC-4, T. Kurt Bond wrote:
> On Saturday, August 12, 2023 at 1:32:27 PM UTC-4, Robert A. Brooks wrote:
>
> > > As cao noticed, the MAP file from the most recent link *is* in the https://github.com/tkurtbond/VTIL repository and the direct link is https://github.com/tkurtbond/VTIL/blob/main/vtil.map
> > >
> > > Sorry, I should have mentioned that.
> > I'm impressed that you somehow crafted a PSECT with a negative length!
> >
> >
> > Psect Name Module Name Base End Length
> > ---------- ----------- ---- --- ------
> >
> > VTILCODE 00002000 00001FFB FFFFFFFC ( 4294967292.)
> > VTIL 00002000 00001FFB FFFFFFFC ( 4294967292.)
> >
> >
> > We'll need John Reagan to weigh in here.
> Well, that doesn't look good. I've updated that repository with the .MAP and VTIL.LIS with /SHOW=EXPANSION from a compilation done today. It shouldn't be too different.

Yay! Pointing out the unusual PSECT size eventually lead me to line 300
in VTIL.MAR:
https://github.com/tkurtbond/VTIL/blob/84496c62b545502961bffd788bbaefdb55450d9e/vtil.mar#L300
Where I saw the assignment "branch = -4". Thaaaat doesn't look right!
Looking at the picture of the source page for that line:
https://github.com/tkurtbond/VTIL-pictures/blob/main/vtil.mar/page-10-lines-285-to-334.jpeg
I see that it should have been "branch = . - 4". That looks better.

And now it links! And when I run it, I get a "VTIL>" prompt! Of course,
nothing seems to work, other than the Ctrl-C handler, and Ctrl-Z to exit.
But now I can run it in the debugger and figure out what is going on.

Thanks for you help!

On Saturday, August 12, 2023 at 1:04:12 PM UTC-7, T. Kurt Bond wrote:
> On Saturday, August 12, 2023 at 3:39:12 PM UTC-4, T. Kurt Bond wrote:
> > On Saturday, August 12, 2023 at 1:32:27 PM UTC-4, Robert A. Brooks wrote:
> >
> > > > As cao noticed, the MAP file from the most recent link *is* in the https://github.com/tkurtbond/VTIL repository and the direct link is https://github.com/tkurtbond/VTIL/blob/main/vtil.map
> > > >
> > > > Sorry, I should have mentioned that.
> > > I'm impressed that you somehow crafted a PSECT with a negative length!
> > >
> > >
> > > Psect Name Module Name Base End Length
> > > ---------- ----------- ---- --- ------
> > >
> > > VTILCODE 00002000 00001FFB FFFFFFFC ( 4294967292.)
> > > VTIL 00002000 00001FFB FFFFFFFC ( 4294967292.)
> > >
> > >
> > > We'll need John Reagan to weigh in here.
> > Well, that doesn't look good. I've updated that repository with the .MAP and VTIL.LIS with /SHOW=EXPANSION from a compilation done today. It shouldn't be too different.
> Yay! Pointing out the unusual PSECT size eventually lead me to line 300
> in VTIL.MAR:
> https://github.com/tkurtbond/VTIL/blob/84496c62b545502961bffd788bbaefdb55450d9e/vtil.mar#L300
> Where I saw the assignment "branch = -4". Thaaaat doesn't look right!
> Looking at the picture of the source page for that line:
> https://github.com/tkurtbond/VTIL-pictures/blob/main/vtil.mar/page-10-lines-285-to-334.jpeg
> I see that it should have been "branch = . - 4". That looks better.
>
> And now it links! And when I run it, I get a "VTIL>" prompt! Of course,
> nothing seems to work, other than the Ctrl-C handler, and Ctrl-Z to exit.
> But now I can run it in the debugger and figure out what is going on.
>
> Thanks for you help!

I see I'm too late for this party. Glad you found the issue, and kudos to Robert Brooks for catching the negative PSECT size!

Best wishes from an old friend you've never met.

Ehud

On 2023-08-12, T. Kurt Bond <tkurtbond@gmail.com> wrote:
>
> Yay! Pointing out the unusual PSECT size eventually lead me to line 300
> in VTIL.MAR:
> https://github.com/tkurtbond/VTIL/blob/84496c62b545502961bffd788bbaefdb55450d9e/vtil.mar#L300
> Where I saw the assignment "branch = -4". Thaaaat doesn't look right!
> Looking at the picture of the source page for that line:
> https://github.com/tkurtbond/VTIL-pictures/blob/main/vtil.mar/page-10-lines-285-to-334.jpeg
> I see that it should have been "branch = . - 4". That looks better.
>

On the plus side, at least you didn't miss out both the "." and the "-"
so at least the linker warned you about it. :-)

BTW, I've never seen that linker error before that I can recall...

Simon.

--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Walking destinations on a map are further away than they appear.

On 2023-08-12, Robert A. Brooks <FIRST.LAST@vmssoftware.com> wrote:
> On 8/12/2023 11:11 AM, T. Kurt Bond wrote:
>> On Saturday, August 12, 2023 at 6:46:11?AM UTC-4, cao[...] wrote:
>>> On Saturday, August 12, 2023 at 2:30:01?AM UTC+1, gah4 wrote:
>>>> Since you have LINK/MAP, you could post the link map.
>>>>
>>>> It looks like an external symbol problem of some kind, where the linker is trying to modify an address, but the wrong place.
>>>>
>>>> Interesting, it is just before what seems to be the beginning.
>>> There's something weird in the map output anyway.
>>> The cluster listings don't make sense and there seem to be a lot of rows for the default cluster.
>>> VTILCODE is also being listed as 4Gb in size roughly.
>>> I noticed that the beginning of VTILCODE, the initial macro expansion has an offset of -4 and 00001FFC is suspiciously 4 away from 00002000
>>> Possibly barking up the wrong tree but I'll have a proper look later.
>>>
>>> K
>>
>> As cao noticed, the MAP file from the most recent link *is* in the https://github.com/tkurtbond/VTIL repository and the direct link is https://github.com/tkurtbond/VTIL/blob/main/vtil.map
>>
>> Sorry, I should have mentioned that.
>
> I'm impressed that you somehow crafted a PSECT with a negative length!
>
>
> Psect Name Module Name Base End Length
> ---------- ----------- ---- --- ------
>
> VTILCODE 00002000 00001FFB FFFFFFFC ( 4294967292.)
> VTIL 00002000 00001FFB FFFFFFFC ( 4294967292.)
>
>
> We'll need John Reagan to weigh in here.
>

On the surface, given the End is less than the Base, it looks like
the kind of thing that should be marked with an explicit error, instead
of the more indirect error the linker currently gives.

Simon.

--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Walking destinations on a map are further away than they appear.

On 8/14/2023 8:47 AM, Simon Clubley wrote:
> On 2023-08-12, Robert A. Brooks <FIRST.LAST@vmssoftware.com> wrote:
>> On 8/12/2023 11:11 AM, T. Kurt Bond wrote:
>>> On Saturday, August 12, 2023 at 6:46:11?AM UTC-4, cao[...] wrote:
>>>> On Saturday, August 12, 2023 at 2:30:01?AM UTC+1, gah4 wrote:
>>>>> Since you have LINK/MAP, you could post the link map.
>>>>>
>>>>> It looks like an external symbol problem of some kind, where the linker is trying to modify an address, but the wrong place.
>>>>>
>>>>> Interesting, it is just before what seems to be the beginning.
>>>> There's something weird in the map output anyway.
>>>> The cluster listings don't make sense and there seem to be a lot of rows for the default cluster.
>>>> VTILCODE is also being listed as 4Gb in size roughly.
>>>> I noticed that the beginning of VTILCODE, the initial macro expansion has an offset of -4 and 00001FFC is suspiciously 4 away from 00002000
>>>> Possibly barking up the wrong tree but I'll have a proper look later.
>>>>
>>>> K
>>>
>>> As cao noticed, the MAP file from the most recent link *is* in the https://github.com/tkurtbond/VTIL repository and the direct link is https://github.com/tkurtbond/VTIL/blob/main/vtil.map
>>>
>>> Sorry, I should have mentioned that.
>>
>> I'm impressed that you somehow crafted a PSECT with a negative length!
>>
>>
>> Psect Name Module Name Base End Length
>> ---------- ----------- ---- --- ------
>>
>> VTILCODE 00002000 00001FFB FFFFFFFC ( 4294967292.)
>> VTIL 00002000 00001FFB FFFFFFFC ( 4294967292.)
>>
>>
>> We'll need John Reagan to weigh in here.
>>
>
> On the surface, given the End is less than the Base, it looks like
> the kind of thing that should be marked with an explicit error, instead
> of the more indirect error the linker currently gives.
>
> Simon.
>

Any language can have some helpful protections/reporting, or not. Assembly on
any system usually has very few. That means that any assembly language coders
had better be rather good at reading their code, and seeing possible errors.

There will always be some assembly language. The less the better.

I think Robert's comment about a negative length PSECT is both funny, and all
that's needed to point out the issue. Good catch.

--
David Froble Tel: 724-529-0450
Dave Froble Enterprises, Inc. E-Mail: davef@tsoft-inc.com
DFE Ultralights, Inc.
170 Grimplin Road
Vanderbilt, PA 15486

On 8/14/2023 12:47 AM, Ehud Gavron wrote:

> Best wishes from an old friend you've never met.

As an old-timer, I remember you from DECUS symposia

--

--- Rob

On 2023-08-14 04:47:10 +0000, Ehud Gavron said:

> Best wishes from an old friend you've never met.
>
> Ehud

Hello from an old friend you HAVE met. I hope you're well and I'm happy
to see you contributing in c.o.v. again.