TUHS

tuhs@tuhs.org

6420 discussions

by Dennis Ritchie

Helbig observed > while wondering how unsigned integer division is implemented in C > I found a bug in V7: a/b and a%b with b >= 2^15 does not give the expected > results, if a and b are unsigned int's. Does http://www.cs.bell-labs.com/who/dmr/odd.html#muldiv suffice to explain the behavior? Dennis

22 years

unsigned division in V7's C

by Wolfgang Helbig

Hi, while wondering how unsigned integer division is implemented in C I found a bug in V7: a/b and a%b with b >= 2^15 does not give the expected results, if a and b are unsigned int's. Was this bug ever noticed or even fixed? Greetings, Wolfgang Here is a program showing the bug: main() { unsigned int a, b; a = 60000; b = 40000; printf("a/b: %u, a%%b: %u\n", a/b, a%b); b = 25000; printf("a/b: %u, a%%b: %u\n", a/b, a%b); } The above program prints a/b: 65534, a%b: 8928 a/b: 2, a%b: 10000 The first line should be of course a/b: 1, a%b: 20000

22 years

Re: Patches to improve 6th Edition

by norman＠nose.cs.utoronto.ca

A belated rejoinder to: This bug existed from (at least) the Fifth up until the Eighth editions from 1127 (BTL research). The Ninth edition had much rewrites in the kernel and the algorthim was changed at that point. You're welcome. I can't remember, though, whether I fixed it on purpose or as part of the many general cleanups and restructurings I did at that point. And I didn't do enough of it; recently I discovered a dumb problem that I probably introduced in the same cleanups, wherein the i-node allocated for a file that couldn't be created because the file system was full and the directory entry couldn't be written was left allocated but nameless ... Norman Wilson Toronto ON

22 years

Re: [pups] Mail delivery failed: returning message to sender

by pete＠dunnington.u-net.com

On Jan 6, 19:01, Wolfgang Helbig wrote: > Richard L. Hamilton searches: > >The files > > > >http://minnie.tuhs.org/UnixTree/V7/usr/sys/dev/rl.c > >http://minnie.tuhs.org/UnixTree/V7/usr/sys/conf/mkconf.c > > > >don't seem to match - mkconf.c wants to generate references to > >rlopen() and rlclose(), which don't exist in that version of rl. > >What I'm really looking for is the right version of driver(s) etc. > > It seems that a version of the rl driver that matches > mkconf, was never supplied by the bell labs. > > As a workaround delete the rl entry in cdevsw and the line > "int rlopen,..." from c.c. This will give you a block device of the rl disk, > but no raw device. Hmm... my 7th Edition distribution /usr/sys/dev/rl.c certainly *does* have rlopen() and rlclose(). If you want them, I can send rl.c and the matching mkconf.c to you, or to Warren, along with the rl1conf and rl2conf (for RL01 and RL02 disks) from /usr/sys/conf/ and any other files needed (if someone can tell me what might be required -- it's a long long time since I looked at any of this. The distribution came for an 11/23 with 2 x RL02, and it also has some special source files for 11/23, 11/34, etc ("small machines") including tests for recovery from segmentation traps in floating point (which isn't the same on all PDP-11 models), and an RK driver from the Boston Children's Museum which handles overlapped seeks on multiple drives. I'll be away for a couple of days so don't expect a prompt reply, but the source is online here so I can easily copy a few files when I get back. -- Pete Peter Turnbull Network Manager University of York

22 years, 1 month

[pups] Mail delivery failed: returning message to sender

by Wolfgang Helbig

Richard L. Hamilton searches: >The files > >http://minnie.tuhs.org/UnixTree/V7/usr/sys/dev/rl.c >http://minnie.tuhs.org/UnixTree/V7/usr/sys/conf/mkconf.c > >don't seem to match - mkconf.c wants to generate references to >rlopen() and rlclose(), which don't exist in that version of rl. >What I'm really looking for is the right version of driver(s) etc. It seems that a version of the rl driver that matches mkconf, was never supplied by the bell labs. As a workaround delete the rl entry in cdevsw and the line "int rlopen,..." from c.c. This will give you a block device of the rl disk, but no raw device. Greetings, Wolfgang

22 years, 1 month

[pups] V7 RL02 driver questions

by Richard L. Hamilton

The files http://minnie.tuhs.org/UnixTree/V7/usr/sys/dev/rl.c http://minnie.tuhs.org/UnixTree/V7/usr/sys/conf/mkconf.c don't seem to match - mkconf.c wants to generate references to rlopen() and rlclose(), which don't exist in that version of rl.c. What I'm really looking for is the right version of driver(s) etc. so that I can regenerate the rl2unix kernel in the V7 image available for use with simh, so that I can add or tweak drivers, eventually adding the dz driver so I can run multiple "terminals" with telnet access. -- mailto:rlhamil@mindwarp.smart.net http://www.smart.net/~rlhamil

22 years, 1 month

db(I) in Unix V6

by Wolfgang Helbig

Hi, tinkering at V6--now I need db(I). It works fine, but three features that are described in its manual page seem to be missing: - The names of the general registers (r0, ...) are not build in, at least db only responds with `?' to `r0'. I can examine the registers via the `$' command though. - The '*' feature doesn't work, i.e. `*_main' gives me a `?'. - I can't specify a number being relocatable, i.e. '10r' gives me a `?'. Is this an exercise to the user? Did anyone solve the exercise. Or did I just not understand the manual page? Greetings, Wolfgang

22 years, 1 month

Re: [pups] V7 setjmp/longjmp

by Wolfgang Helbig

Michael Davidson distrusted: >Note that this means that r4, r3 and r2 get restored to the values >that they had at the time that the next stack frame was established >(ie when the "next" function down was called from the original context) >and that these are not necessarily the values that they had at the >moment when the original call to setjnp returned. Hence the various >caveats about not relying on the value of any register variables >after returning via longjmp. You can trust register variables when setjmp() returns the second time. Then the registers are restored to the values they had when the "next" function was called, that is the "values as of the time longjmp() was called" (quoted from longjmp(3)'s man page. Thus any variable behaves the same, regardless of its storage class. By the way, this is not the case in V6. The register variables-- and only the register variables-- are restored to the values as of the time setexit() was called. (In V6 "setjmp()" is named "setexit()" and "longjmp()" is named "reset()"). Thus the value of a variable depends on its storage class. If it is register, you get the value when setexit() was called, if it is not register, you get the value when reset() was called. This might be considered a bug. Greetings Wolfgang

22 years, 1 month

[pups] V7 setjmp/longjmp

by bwc＠coraid.com

Looking thru the code for setjmp/longjmp (/usr/src/libc/gen/setjmp.s), and have a question. Longjmp seems to make an attempt at restoring register two through four, but setjmp doesn't save them. Where do these register get saved from? Brantley Coile

22 years, 1 month

Re: Patches to improve 6th Edition

by Wolfgang Helbig

some hours ago I wrote: >I ask this, because my system crashes ones in a while. But I changed it >here and there, so it might be my fault. It *was* my fault! So far, I have only found one bug in V6: The size of swapmap and coremap are configured too small, so they might overflow. The sizes don't take account of shared text segments, that produce extra chunks of free memory. In V6 there are 19 pure executables, so their minimal size should be (NPROC+19)*2, or better yet (NPROC+NTEXT)*2, where NTEXT should be set to 20 instead of 40 as distributed in param.h . I don't think this bug ever showed up. It is very unlikely. In V7 these sizes are choosen to be even smaller than NPROC and even with its many pure executables, I don't think V7 ever crashed because of this bug. So, with the only exception reported by Brian S. Walden, there seem to be no bugs in V6 that crash systems. Greetings Wolfgang

22 years, 1 month

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

TUHS