[comp.unix.ultrix] Ultrix 3.1C /bin/csh looping to memory limit

D. Allen [CGL]) (09/24/90)

Could someone with a DECsystem running Ultrix 4.0 try this for me?
This is what happens on my DECsystem 5400 running Ultrix 3.1C.

    % /bin/csh -f      # don't do this to your login shell
    % cat >/tmp/foo
    echo before
    echo $nosuchvar
    ^D
    % limit data 2m    # for safety!
    % source /tmp/foo
    before
    nosuchvar: Undefined variable
    data size rlimit exceeded, pid 259, proc csh
    Out of memory.

"^D" is your EOF character.

This bug is much more exciting when you have your datasize limit up
around 90Mb.  The dying shell throws everyone else out of memory before
it eeks out a horrible death.  Grinds the system to a halt for a minute
or two.  This does it too:

    % cat >/tmp/foo
    echo before
    echo "$<"
    ^D
    % limit data 2m    # for safety!
    % source /tmp/foo
    before
    ^C
    data size rlimit exceeded, pid 226, proc csh
    Out of memory.

"^C" is your interrupt character.  You'll have to push <CR> after your
interrupt character, since C Shells incorrectly hold interrupts when
reading from $<.

In fact, anything that causes CSH to start error recovery in a sourced
file will trigger a loop that will eat all the memory on your machine.

If you prefer pure CPU loops, just do this:

    % limit cpu 5
    % source /dev/tty
    % echo nosuchfile*
    echo: No match.

That shell will loop until it dies exceeding the CPU limit.

Is this fixed in DEC MIPS Ultrix 4.0?
-- 
-IAN! (Ian! D. Allen) idallen@watcgl.uwaterloo.ca idallen@watcgl.waterloo.edu
 [129.97.128.64]  Computer Graphics Lab/University of Waterloo/Ontario/Canada

grr@cbmvax.commodore.com (George Robbins) (09/24/90)

In article <1990Sep23.211812.1479@watcgl.waterloo.edu> idallen@watcgl.waterloo.edu (Ian! D. Allen [CGL]) writes:
> Could someone with a DECsystem running Ultrix 4.0 try this for me?
> This is what happens on my DECsystem 5400 running Ultrix 3.1C.

I'm running the 4.0 ultrix csh under 3.1C and the csh looping / memory muncher
problems seem to be fixed.  Note that they should also be fixed if you installed
the 3.1C Mandatory Patch Tape, albeit at the loss of the command completion
features.

If you haven't installed this tape, you are asking for problems...

-- 
George Robbins - now working for,     uucp:   {uunet|pyramid|rutgers}!cbmvax!grr
but no way officially representing:   domain: grr@cbmvax.commodore.com
Commodore, Engineering Department     phone:  215-431-9349 (only by moonlite)

frank@croton.enet.dec.com (Frank Wortner) (09/24/90)

> Could someone with a DECsystem running Ultrix 4.0 try this for me?

Done:

$ /bin/csh -r
csh> cat >/tmp/foo
echo before
echo $nosuchvar
csh> limit data 2m
csh> source /tmp/foo
before
nosuchvar: Undefined variable.
csh>

Fixed in 4.0!

					Frank

jmg@cernvax.UUCP (mike gerard) (09/24/90)

In article <14603@cbmvax.commodore.com> grr@cbmvax.commodore.com (George Robbins) writes:
>I'm running the 4.0 ultrix csh under 3.1C and the csh looping / memory muncher
>problems seem to be fixed.  Note that they should also be fixed if you installed
>the 3.1C Mandatory Patch Tape, albeit at the loss of the command completion
>features.
>
>If you haven't installed this tape, you are asking for problems...

Having many times tried to find out what patches exist for Ultrix
AND
having hit a brick wall
my question is:-

Is it mandatory to be told about the existence of this patch (and any
other similar ones)?

(and does it depend upon whether you have a maintenance contract and/or
a media&documentation update contract?)
-- 
 _ _  o |            __                    |    jmg@cernvax.uucp
| | |   |     _     /  \  _   __  _   __  _|    jmg@cernvax.bitnet
| | | | |_)  /_)    |  __/_) | (___\ | (_/ |  J. M. Gerard, Div. DD, CERN,
| | |_|_| \_/\___   \__/ \___|   (_|_|   \_|_ 1211 Geneva 23, Switzerland

grr@cbmvax.commodore.com (George Robbins) (09/25/90)

In article <2815@cernvax.UUCP> jmg@cernvax.UUCP (mike gerard) writes:
> In article <14603@cbmvax.commodore.com> grr@cbmvax.commodore.com (George Robbins) writes:
> >I'm running the 4.0 ultrix csh under 3.1C and the csh looping / memory muncher
> >problems seem to be fixed.  Note that they should also be fixed if you installed
> >the 3.1C Mandatory Patch Tape, albeit at the loss of the command completion
> >features.
> >
> >If you haven't installed this tape, you are asking for problems...
> 
> Having many times tried to find out what patches exist for Ultrix
> AND
> having hit a brick wall
> my question is:-
> 
> Is it mandatory to be told about the existence of this patch (and any
> other similar ones)?
> 
> (and does it depend upon whether you have a maintenance contract and/or
> a media&documentation update contract?)

I suppose only someone from DEC can really answer this (Fred?, Kurt?) and
the answer may vary depending on whether you're in the US or elsewhere.

For history's sake, Ultrix 3.1C was released somewhere around late December 89,
to provide support for the DECsystem 5800.  There were a number of bugs of the
instant crash and burn variety, at least on some 5800 configurations.
In February, (yes, it was a loooong wait) DEC finally released a "Mandatory
Patch Tape" that cleaned up the worst of these problems.

For reference, the patch letter reference order number  AV-PBPAA-TE and
the tape (TK50) is marked AQ-PBPBA-BE.

My interpretation of "Mandatory Patch Tape" is that without it, the product
doesn't function/conform to the description on the SPD, and that, in effect,
it is a warranty remedy distributed to all who received the original 3.1C
distribution.  (this interpretation may be completely bogus).

I haven't received any new 3.1C distributions, so I don't know what the
current arrangement is.  Traditionally, the patch tape would be included
in the distribution kit, however it is also possible that the patches
were folded into a new "edition" of the 3.1C release tape.

Also, as far as I know, there weren't any similar "Mandatory Patch Tapes"
for Ultrix 3.1/3.1A/3.1B, since a lot of the problems were 5810 related.
In this case all you can do is the normal fishing with the support center
to get the appropriate patches for your release.  Ordering 3.1D media and
documentation is probably a better idea if you aren't ready to brave 4.0.

The following problems were fixed by this tape, you can probably assume
there are individual patches available if the appear in other releases...

problems fixed:

DMB port != 0 instant crash
csh eats all of memory from time to time
quota check seg fault
missing core files on seg fault
mt status seg fault
decnet problems with if_ln
nfs causes panic tblmis
nfs locking performance problems
13 LAT related problems
  (including panic with LAT printers)

release notes:

warning about disconnected lines
warning about removing "QUOTA" from config file
note that non-supported materials are on UWS V2.0 tape
cpio seg fault
note about debni device name.

-- 
George Robbins - now working for,     uucp:   {uunet|pyramid|rutgers}!cbmvax!grr
but no way officially representing:   domain: grr@cbmvax.commodore.com
Commodore, Engineering Department     phone:  215-431-9349 (only by moonlite)

grr@cbmvax.commodore.com (George Robbins) (09/25/90)

In article <14628@cbmvax.commodore.com> grr@cbmvax.commodore.com (George Robbins) writes:
> 
> For reference, the patch letter reference order number  AV-PBPAA-TE and
> the tape (TK50) is marked AQ-PBPBA-BE.

Actually the order number for the kit appears to be QA-VVTAN-T53.1P,
titled "ULT-32 V3.1C TK50 UPD"....

I'd suggest you give some DECperson a day or two to respond before trying
to order this, the information may be grossly out of date...

-- 
George Robbins - now working for,     uucp:   {uunet|pyramid|rutgers}!cbmvax!grr
but no way officially representing:   domain: grr@cbmvax.commodore.com
Commodore, Engineering Department     phone:  215-431-9349 (only by moonlite)