Re: [TUHS] changes in C compilers

Re: [TUHS] changes in C compilers

[Norman Wilson]

Clem Cole:

  Clearly from the time, ditroff did not yet exist.   The more I think about
  it, Brian K actually might know some of the story.

===

I'm quite sure Brian Kernighan knows the full story of the origins
of typesetter-independent troff (as it was originally called, in
CSTR 97, published in 1981; the binary was just /usr/bin/troff).

The reason I'm so sure of that is that it was Brian who rewrote
troff to bring it into the modern era and to make it supportable.
He's also the author of the CSTR.

Norman Wilson
Toronto ON

Re: [TUHS] changes in C compilers

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 591 bytes --]

On Thu, Aug 30, 2018 at 4:27 PM Norman Wilson <norman@oclsc.org> wrote:

> Clem Cole:
>
>   Clearly from the time, ditroff did not yet exist.   The more I think
> about
>   it, Brian K actually might know some of the story.
>
> ===
>
> I'm quite sure Brian Kernighan knows the full story of the origins

That was not what I was saying....  (yes, I know Brian wrote ditroff and
have spoken at length to him about actually).
What, I was saying since he was Brian Ried's advisor on Scribe (during that
timeframe BTW), he might know some of the other details I refered too.
ᐧ

[-- Attachment #2: Type: text/html, Size: 1496 bytes --]

Re: [TUHS] changes in C compilers

[Noel Chiappa]

    > From: Clem Cole

    > Looking at the v6 distribution tape I have, the assembler versions of
    > roff and nroff was there
    
Whoa! The standard V6 distribution tape, as in the one there are a couple of
copies of in the repository, does not have that.

Do you have that in machine-readable form? If so, can you get it to Mr. Toomey
ASAP?


    > The order I remember is this ... V5, V6, Patches, Typesetter C, TS, V7

Where do USG and PWB fit into that?

The repository has PWB (or, what is _claimed_ to be PWB), which is how I know
the MIT system is PWB. But there is nothing of the others (except the kernel
manual for USG, which shows that the version described in it is basically V6).

If anyone has TS in _any_ form (including hardcopy listings, please speak up!
Those 'early' PDP-11 versions are very poorly documented now, and I'd love to
get more on them.

    Noel

Re: [TUHS] changes in C compilers

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 3099 bytes --]

On Sat, Aug 25, 2018 at 7:14 PM Noel Chiappa <jnc@mercury.lcs.mit.edu>
wrote:

>     > From: Clem Cole
>
>     > Looking at the v6 distribution tape I have, the assembler versions of
>     > roff and nroff was there
>
> Whoa! The standard V6 distribution tape, as in the one there are a couple
> of
> copies of in the repository, does not have that.

The v6 stuff is in Warrens archives.  The files are in the ./s7 directory:
{,n}roff[1-8].s  show dates of July 17, 1975


>
> Do you have that in machine-readable form? If so, can you get it to Mr.
> Toomey
> ASAP?
>
>
>     > The order I remember is this ... V5, V6, Patches, Typesetter C, TS,
> V7
>
> Where do USG and PWB fit into that?

This is order of the releases  we had at CMU.


PWB 1.0 was before USG was formed I want to say '77 Mashey lead the team
for sure and I think Frazer may have been part.

USG was Created and APS and Ted were two of First hackers who were there.
Ted tried to get me at some point but I did not want to move to NJ.

Fwiw: The job of USG was to support Unix for the operating companies that
wanted to use Unix at the TelCos.   Steve Johnson I think went over there
from Research at some point (I assume to lead the development tools).

PWB 2.0 was the first Formal full USG release - I think in 1980 but they
did stuff before that.  steve might remember who the original managers were
And other things they released -  I thought Mashey had moved on by then.
Korn's stuff became the Core of the toolkit releases at some point but
dates on that are fuzzy in my memory.

also, Until PWB 3.0 (aka System III) none of the USG releases were licensed
to universities or anyone outside the Bell System . .   but a number of
at&t employees brought it with them On their OYOC times so 1.0, 2.0 etc all
leaked.

I thought TS was done in MH not Summit.  Heinz L was doing Merit around
that time too.    As I understand it, the TS kernel was attempt at bringing
the PWB 1.0 changes in line with what Research had at the time.  Some later
version of That kernel plus some of the Columbus changes for things like
shared memory and ipc became either PWB 2.0/3.0.

PWB 4.0 went to the TelCos but was not released outside.  I don't know of
any University that was interested in it because by the BSD was available
so even if an employee brought it, I don't think many people wanted it.





>
> The repository has PWB (or, what is _claimed_ to be PWB), which is how I
> know
> the MIT system is PWB. But there is nothing of the others (except the
> kernel
> manual for USG, which shows that the version described in it is basically
> V6)


Right that was leaked somehow.


.
>
> If anyone has TS in _any_ form (including hardcopy listings, please speak
> up!

I have looked in fact the week I did a huge clean out in the basement.  So
far nothing new. I'll let you know if I find anything readable.



> Those 'early' PDP-11 versions are very poorly documented now, and I'd love
> to
> get more on them.



Yeah as I said it was stream with a lot of cross pollination.

Clem
-- 
Sent from a handheld expect more typos than usual

[-- Attachment #2: Type: text/html, Size: 4861 bytes --]

[TUHS] changes in C compilers

[arnold]

My two cents, ...

> From: Clem Cole <clemc@ccc.com>
> Date: Thu, 23 Aug 2018 20:30:19 -0400
> To: ron@ronnatalie.com
> Subject: Re: [TUHS] C++ / Kernel
>
> Yep.  Im pretty sure I remember void being in typesetter C also.   IIRC the
> differences between that version of Dennis???s compiler and what was included
> in 7th Edition was mostly in the libraries ie stdio was first released as
> part of the typesetter compiler but it was still a work in progress.

K&R 1 did not have void or structure assignment. Those came later,
although I'm not sure when.  They may have been mentioned in an
appendix; my copy isn't handy to check.

At what point did each struct become its own namespace? I think
around the time of K&R1.

> From: Clem cole <clemc@ccc.com>
> Date: Thu, 23 Aug 2018 22:52:24 -0400
> To: Noel Chiappa <jnc@mercury.lcs.mit.edu>
> Subject: Re: [TUHS] C++ / Kernel
>
> ...
>
> The big changes to the language were between 6th Edition and Typesetter
> which were done in concert if not to support Brian???s work on the troff
> rewrite.  Plus the first draft of book was being written around then also.

The troff rewrite was later, circa '81 or so. Definitely NOT in the
V6/V7 timeframe.

Arnold

Re: [TUHS] changes in C compilers

[Clem cole]

Arnold, I caution on absolutes.   Typesetter C predates V7.  It was for V6 but a release of ditroff  may have been after V7. - do remember something like that in the early 80s.  What I do not remember the order of which compilers.  The driver for writing ditroff was the Labs getting an APS5 typesetter which was in the late 70s.  

And (I’m less) certain here, I think Bourne once told me they used the APS5 for the V7 manual set masters because it supported more fonts.   

Btw. The whole reason why typesetter C had a compiler included was the v6 compiler was not insufficient (primarily missing support for stdio) but V7 was not.  

That said the dates and some of the ordering are fuzzy in my mind at this point.  I’m trying to use other things that I have absolute dates for and trying line them up.   

We had V5, quickly upgraded the V6, then TS and finally V7.  I left in late Spring ‘79.   That order I know is right and Ted was the reason we had TS as he brought it with him in 76 / 77.  (Same for the proofs for K&R1).   

We got typesetter C at some point I thought before Ted brought TS but that’s fuzzy.  

Sent from my PDP-7 Running UNIX V0 expect things to be almost but not quite. 

> On Aug 24, 2018, at 3:02 AM, arnold@skeeve.com wrote:
> 
> My two cents, ...
> 
>> From: Clem Cole <clemc@ccc.com>
>> Date: Thu, 23 Aug 2018 20:30:19 -0400
>> To: ron@ronnatalie.com
>> Subject: Re: [TUHS] C++ / Kernel
>> 
>> Yep.  Im pretty sure I remember void being in typesetter C also.   IIRC the
>> differences between that version of Dennis???s compiler and what was included
>> in 7th Edition was mostly in the libraries ie stdio was first released as
>> part of the typesetter compiler but it was still a work in progress.
> 
> K&R 1 did not have void or structure assignment. Those came later,
> although I'm not sure when.  They may have been mentioned in an
> appendix; my copy isn't handy to check.
> 
> At what point did each struct become its own namespace? I think
> around the time of K&R1.
> 
>> From: Clem cole <clemc@ccc.com>
>> Date: Thu, 23 Aug 2018 22:52:24 -0400
>> To: Noel Chiappa <jnc@mercury.lcs.mit.edu>
>> Subject: Re: [TUHS] C++ / Kernel
>> 
>> ...
>> 
>> The big changes to the language were between 6th Edition and Typesetter
>> which were done in concert if not to support Brian???s work on the troff
>> rewrite.  Plus the first draft of book was being written around then also.
> 
> The troff rewrite was later, circa '81 or so. Definitely NOT in the
> V6/V7 timeframe.
> 
> Arnold

Re: [TUHS] changes in C compilers

[arnold]

The APS work started in the summer of 1979. See http://www.eprg.org/papers/202paper.pdf
and see some of the other stuff at
http://www.cs.princeton.edu/~bwk/202/index.html.

I think that's after V7 was released.

Arnold

Clem cole <clemc@ccc.com> wrote:

> Arnold, I caution on absolutes.   Typesetter C predates V7.  It was for V6 but a release of ditroff  may have been after V7. - do remember something like that in the early 80s.  What I do not remember the order of which compilers.  The driver for writing ditroff was the Labs getting an APS5 typesetter which was in the late 70s.  
>
> And (I???m less) certain here, I think Bourne once told me they used the APS5 for the V7 manual set masters because it supported more fonts.   
>
> Btw. The whole reason why typesetter C had a compiler included was the v6 compiler was not insufficient (primarily missing support for stdio) but V7 was not.  
>
> That said the dates and some of the ordering are fuzzy in my mind at this point.  I???m trying to use other things that I have absolute dates for and trying line them up.   
>
> We had V5, quickly upgraded the V6, then TS and finally V7.  I left in late Spring ???79.   That order I know is right and Ted was the reason we had TS as he brought it with him in 76 / 77.  (Same for the proofs for K&R1).   
>
> We got typesetter C at some point I thought before Ted brought TS but that???s fuzzy.  
>
> Sent from my PDP-7 Running UNIX V0 expect things to be almost but not quite. 
>
> > On Aug 24, 2018, at 3:02 AM, arnold@skeeve.com wrote:
> > 
> > My two cents, ...
> > 
> >> From: Clem Cole <clemc@ccc.com>
> >> Date: Thu, 23 Aug 2018 20:30:19 -0400
> >> To: ron@ronnatalie.com
> >> Subject: Re: [TUHS] C++ / Kernel
> >> 
> >> Yep.  Im pretty sure I remember void being in typesetter C also.   IIRC the
> >> differences between that version of Dennis???s compiler and what was included
> >> in 7th Edition was mostly in the libraries ie stdio was first released as
> >> part of the typesetter compiler but it was still a work in progress.
> > 
> > K&R 1 did not have void or structure assignment. Those came later,
> > although I'm not sure when.  They may have been mentioned in an
> > appendix; my copy isn't handy to check.
> > 
> > At what point did each struct become its own namespace? I think
> > around the time of K&R1.
> > 
> >> From: Clem cole <clemc@ccc.com>
> >> Date: Thu, 23 Aug 2018 22:52:24 -0400
> >> To: Noel Chiappa <jnc@mercury.lcs.mit.edu>
> >> Subject: Re: [TUHS] C++ / Kernel
> >> 
> >> ...
> >> 
> >> The big changes to the language were between 6th Edition and Typesetter
> >> which were done in concert if not to support Brian???s work on the troff
> >> rewrite.  Plus the first draft of book was being written around then also.
> > 
> > The troff rewrite was later, circa '81 or so. Definitely NOT in the
> > V6/V7 timeframe.
> > 
> > Arnold

Re: [TUHS] changes in C compilers

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 3407 bytes --]

On Sat, Aug 25, 2018 at 2:38 PM <arnold@skeeve.com> wrote:

> The APS work started in the summer of 1979. See
> http://www.eprg.org/papers/202paper.pdf
> and see some of the other stuff at
> http://www.cs.princeton.edu/~bwk/202/index.html.
>
> I think that's after V7 was released.
>
Ok, so that was clearly the first ditroff.

Typesetter C *must have been the original troff release* which was separate
from V6; but I don't remember what all was in the release.   Looking at the
v6 distribution tape I have, the assembler versions of roff and nroff was
there; but not troff.   V7 clearly shows the original troff in the sources.


The order I remember is this ... V5, V6, Patches, Typesetter C, TS, V7 ...
 although TS and Typesetter might be switched but I know we got Typesetter
C before we got V7.    Ted brought TS to us (in EE) and I thought that had
the new compiler.   CS got TS from us in EE.    But somebody at CMU had
wanted troff because we had the XGP in CS that we drive with Scribe (I want
to say that was EE but I don't remember who was involved).   So I have
memory of somebody hacking on the compiler at some point.  The POR (which
if ever came to bear at CMU was after I left) was some type of hacking on
troff to support the XGP.  Given the time Aharon points out, it might have
been direct support it or it might have been something like vcat - I was
not involved.   Klone might remember more of that.

Clearly from the time, ditroff did not yet exist.   The more I think about
it, Brian K actually might know some of the story.  Scribe was Brian Reid's
PhD Thesis and Brian K was on Reid's committee at the time and I'm guessing
could somehow have been mixed up.

FWIW: Compiler hacking at CMU stands out in my mind because of the 11/40e
had CSAV/CRET instructions.  The CS versions of the compilers generated
code using that, because they had 11/40e with CMU WCS options.  The rest of
us in EE, BioMed, Mellon Institute etc were running on 11/34's or 11/34A
which could not handle those binaries (no WCS).   So I personally spent
time tracking the CS versions of the compiler and bringing things to EE,
trying to keep thing clean.  That was one of my jobs at the time.

That's fairly sure of the order, because we had Typesetter C at CMU in the
Summer '78 when were we negotiating the 'university' commercial V7 license
with Al Arms [which I was personally mixed up -- the finally
ruling/agreement was license one system as a commercial system at the $20K
fee and a university, could then use UNIX for back office and commercial
style uses like Industry.  Al did not require the $5K second CPU stuff from
the Universities, if they got a single $20K license; everyone was happy -
details off list or another thread if you want them; although I will say
CMU was first in early '79, followed by Case in late 1979].

So again, I try to date by things I know are fixed in time and then work
from there.   As Dan points out the cross pollination was high in those
days and it was not just from the labs to the Universities.  For instance,
Ted took fsck back to Summit & MH, as well as a number of other tools
(although I think that one had the longest reach).   Noel has mentioned
similar stories from MIT.  Chesson brought all the networking stuff from
UoI and we saw some of it in datakit (an earlier version of his mpx code
for V7 he did as a grad student).   You get the idea....

Clem

[-- Attachment #2: Type: text/html, Size: 5467 bytes --]

Re: [TUHS] changes in C compilers

[John P. Linderman]

[-- Attachment #1: Type: text/plain, Size: 4128 bytes --]

*Ted took fsck back to Summit & MH*

I got my introduction to UNIX in '73 or '74, when the group running the
11/45 in Piscataway found out I came into work before 6am. UNIX was so
unstable back then that it had to be rebooted every day, to contain file
system corruption. A 6 am reboot went pretty much unnoticed. I could swear
we ran something very like fsck after each reboot. In particular, I recall
the **gok** diagnostic when the type of an inode wasn't anything
recognizable. Whatever we ran, I'm sure it continued to evolve.


On Sat, Aug 25, 2018 at 3:58 PM, Clem Cole <clemc@ccc.com> wrote:

>
>
> On Sat, Aug 25, 2018 at 2:38 PM <arnold@skeeve.com> wrote:
>
>> The APS work started in the summer of 1979. See
>> http://www.eprg.org/papers/202paper.pdf
>> and see some of the other stuff at
>> http://www.cs.princeton.edu/~bwk/202/index.html.
>>
>> I think that's after V7 was released.
>>
> Ok, so that was clearly the first ditroff.
>
> Typesetter C *must have been the original troff release* which was
> separate from V6; but I don't remember what all was in the release.
>  Looking at the v6 distribution tape I have, the assembler versions of roff
> and nroff was there; but not troff.   V7 clearly shows the original troff
> in the sources.
>
>
> The order I remember is this ... V5, V6, Patches, Typesetter C, TS, V7
> ...   although TS and Typesetter might be switched but I know we got
> Typesetter C before we got V7.    Ted brought TS to us (in EE) and I
> thought that had the new compiler.   CS got TS from us in EE.    But
> somebody at CMU had wanted troff because we had the XGP in CS that we drive
> with Scribe (I want to say that was EE but I don't remember who was
> involved).   So I have memory of somebody hacking on the compiler at some
> point.  The POR (which if ever came to bear at CMU was after I left) was
> some type of hacking on troff to support the XGP.  Given the time Aharon
> points out, it might have been direct support it or it might have been
> something like vcat - I was not involved.   Klone might remember more of
> that.
>
> Clearly from the time, ditroff did not yet exist.   The more I think about
> it, Brian K actually might know some of the story.  Scribe was Brian Reid's
> PhD Thesis and Brian K was on Reid's committee at the time and I'm guessing
> could somehow have been mixed up.
>
> FWIW: Compiler hacking at CMU stands out in my mind because of the 11/40e
> had CSAV/CRET instructions.  The CS versions of the compilers generated
> code using that, because they had 11/40e with CMU WCS options.  The rest of
> us in EE, BioMed, Mellon Institute etc were running on 11/34's or 11/34A
> which could not handle those binaries (no WCS).   So I personally spent
> time tracking the CS versions of the compiler and bringing things to EE,
> trying to keep thing clean.  That was one of my jobs at the time.
>
> That's fairly sure of the order, because we had Typesetter C at CMU in the
> Summer '78 when were we negotiating the 'university' commercial V7 license
> with Al Arms [which I was personally mixed up -- the finally
> ruling/agreement was license one system as a commercial system at the $20K
> fee and a university, could then use UNIX for back office and commercial
> style uses like Industry.  Al did not require the $5K second CPU stuff from
> the Universities, if they got a single $20K license; everyone was happy -
> details off list or another thread if you want them; although I will say
> CMU was first in early '79, followed by Case in late 1979].
>
> So again, I try to date by things I know are fixed in time and then work
> from there.   As Dan points out the cross pollination was high in those
> days and it was not just from the labs to the Universities.  For instance,
> Ted took fsck back to Summit & MH, as well as a number of other tools
> (although I think that one had the longest reach).   Noel has mentioned
> similar stories from MIT.  Chesson brought all the networking stuff from
> UoI and we saw some of it in datakit (an earlier version of his mpx code
> for V7 he did as a grad student).   You get the idea....
>
> Clem
>
>

[-- Attachment #2: Type: text/html, Size: 7170 bytes --]

Re: [TUHS] changes in C compilers

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 5114 bytes --]

Fsck was very much a CMU program. I had a small hand in it - the reason why
the errors Messages are upper case in fact. (a different story).    That
said it was hardly a new idea - Ted modelled it from a similar program he
saw at UMich on MTS and one I ran on TSS called the file system checker.
 im guessing there may have been other programs like it.

Before fsck Ken had icheck ncheck and dcheck.  IIRC Masheys team in PWB @
Whippany I think, wrote fsdb which shows up around the same time     I also
think the Columbus folks may have something early on also.

Ted took fsck back To USG and it became part of the Summit releases.  Bill
Joy had been his roommate at Michigan which I think is how to made to ucb.
But Joy had it there before I showed up.

Clem


On Sat, Aug 25, 2018 at 6:19 PM John P. Linderman <jpl.jpl@gmail.com> wrote:

> *Ted took fsck back to Summit & MH*
>
> I got my introduction to UNIX in '73 or '74, when the group running the
> 11/45 in Piscataway found out I came into work before 6am. UNIX was so
> unstable back then that it had to be rebooted every day, to contain file
> system corruption. A 6 am reboot went pretty much unnoticed. I could swear
> we ran something very like fsck after each reboot. In particular, I recall
> the **gok** diagnostic when the type of an inode wasn't anything
> recognizable. Whatever we ran, I'm sure it continued to evolve.
>
>
> On Sat, Aug 25, 2018 at 3:58 PM, Clem Cole <clemc@ccc.com> wrote:
>
>>
>>
>> On Sat, Aug 25, 2018 at 2:38 PM <arnold@skeeve.com> wrote:
>>
>>> The APS work started in the summer of 1979. See
>>> http://www.eprg.org/papers/202paper.pdf
>>> and see some of the other stuff at
>>> http://www.cs.princeton.edu/~bwk/202/index.html.
>>>
>>> I think that's after V7 was released.
>>>
>> Ok, so that was clearly the first ditroff.
>>
>> Typesetter C *must have been the original troff release* which was
>> separate from V6; but I don't remember what all was in the release.
>>  Looking at the v6 distribution tape I have, the assembler versions of roff
>> and nroff was there; but not troff.   V7 clearly shows the original troff
>> in the sources.
>>
>>
>> The order I remember is this ... V5, V6, Patches, Typesetter C, TS, V7
>> ...   although TS and Typesetter might be switched but I know we got
>> Typesetter C before we got V7.    Ted brought TS to us (in EE) and I
>> thought that had the new compiler.   CS got TS from us in EE.    But
>> somebody at CMU had wanted troff because we had the XGP in CS that we drive
>> with Scribe (I want to say that was EE but I don't remember who was
>> involved).   So I have memory of somebody hacking on the compiler at some
>> point.  The POR (which if ever came to bear at CMU was after I left) was
>> some type of hacking on troff to support the XGP.  Given the time Aharon
>> points out, it might have been direct support it or it might have been
>> something like vcat - I was not involved.   Klone might remember more of
>> that.
>>
>> Clearly from the time, ditroff did not yet exist.   The more I think
>> about it, Brian K actually might know some of the story.  Scribe was Brian
>> Reid's PhD Thesis and Brian K was on Reid's committee at the time and I'm
>> guessing could somehow have been mixed up.
>>
>> FWIW: Compiler hacking at CMU stands out in my mind because of the 11/40e
>> had CSAV/CRET instructions.  The CS versions of the compilers generated
>> code using that, because they had 11/40e with CMU WCS options.  The rest of
>> us in EE, BioMed, Mellon Institute etc were running on 11/34's or 11/34A
>> which could not handle those binaries (no WCS).   So I personally spent
>> time tracking the CS versions of the compiler and bringing things to EE,
>> trying to keep thing clean.  That was one of my jobs at the time.
>>
>> That's fairly sure of the order, because we had Typesetter C at CMU in
>> the Summer '78 when were we negotiating the 'university' commercial V7
>> license with Al Arms [which I was personally mixed up -- the finally
>> ruling/agreement was license one system as a commercial system at the $20K
>> fee and a university, could then use UNIX for back office and commercial
>> style uses like Industry.  Al did not require the $5K second CPU stuff from
>> the Universities, if they got a single $20K license; everyone was happy -
>> details off list or another thread if you want them; although I will say
>> CMU was first in early '79, followed by Case in late 1979].
>>
>> So again, I try to date by things I know are fixed in time and then work
>> from there.   As Dan points out the cross pollination was high in those
>> days and it was not just from the labs to the Universities.  For instance,
>> Ted took fsck back to Summit & MH, as well as a number of other tools
>> (although I think that one had the longest reach).   Noel has mentioned
>> similar stories from MIT.  Chesson brought all the networking stuff from
>> UoI and we saw some of it in datakit (an earlier version of his mpx code
>> for V7 he did as a grad student).   You get the idea....
>>
>> Clem
>>
>>
> --
Sent from a handheld expect more typos than usual

[-- Attachment #2: Type: text/html, Size: 8620 bytes --]