Debian Bug report logs - #24085
acct: binaries do not work on 2.1 kernels

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox):

From: <reiter@forest.bofh.asn.au>
To: submit@bugs.debian.org
Subject: acct: binaries do not work on 2.1 kernels
Date: Tue, 30 Jun 1998 17:07:56 +1000 (EST)

Package: acct
Version: 6.3.5-3

Somewhere in the 2.1 series (I only know that it's before 2.1.70), struct
acct has changed.

The result is that the acct programs (as come in prebuilt binaries)
do not work for 2.1 kernels.  (E.g. lastcomm prints rubbish then segfaults.)

I haven't yet tried building from source.  Looking at the kernel source,
it appears that rebuilding from source wouldn't work either, as 
2.1 uses a sort of floating point representation for some of the numbers.
(See linux/kernel/acct.c, in particular encode_comp_t and acct_process.
In linux-2.0, it's in kernel/sys.c.)  Fortunately, the floating point
format used is such that the numbers aren't far off (IISeeC).

Note that I've also submitted a bug report to the author of the kernel
code and suggested that he might consider changing format again, and also
suggested (independent of such a change) a couple of other changes to acct.h.
Therefore, I suggest that you either wait until include/linux/acct.h changes 
again in the 2.1 series before seriously trying to address the problem
(other than perhaps in the package description and/or testing uname -r during
installation), or else I'll send you a copy of the message I sent to MvW.

-- System Information
Debian Release: 2.0 (frozen)
Kernel Version: Linux forest 2.1.107 #1 Tue Jun 30 10:56:43 EST 1998 i486 unknown

Versions of the packages acct depends on:
libc6	Version: 2.0.7pre3-1
bsdutils	Version: 4.2

Message #10 received at 24085@bugs.debian.org (full text, mbox):

From: Dirk Eddelbuettel <edd@debian.org>
To: <reiter@forest.bofh.asn.au>, 24085@bugs.debian.org
Cc: control@bugs.debian.org
Subject: Re: Bug#24085: acct: binaries do not work on 2.1 kernels
Date: Tue, 30 Jun 1998 23:11:41 -0400 (EDT)

forwarded 24085 Noel Cragg <noel@red-bean.com>
merge 18314 24085
quit

Thanks for the bugreport.  The problem is actually known, and the acct
upstream author already made one attempt at fixing this (with version
6.3.4). Unfortunately, that didn't quite work and had to be backed out again
for version 6.3.5.

I have therefore registered your bugreport as 'forwarded' and merged it with
the previous report on the same topic. 

Thanks, Dirk

  Reiter> Package: acct 
  Reiter> Version: 6.3.5-3
  Reiter> 
  Reiter> Somewhere in the 2.1 series (I only know that it's before 2.1.70),
  Reiter> struct acct has changed.
  Reiter> 
  Reiter> The result is that the acct programs (as come in prebuilt binaries)
  Reiter> do not work for 2.1 kernels.  (E.g. lastcomm prints rubbish then
  Reiter> segfaults.)
  Reiter> 
  Reiter> I haven't yet tried building from source.  Looking at the kernel
  Reiter> source, it appears that rebuilding from source wouldn't work
  Reiter> either, as 2.1 uses a sort of floating point representation for
  Reiter> some of the numbers.  (See linux/kernel/acct.c, in particular
  Reiter> encode_comp_t and acct_process.  In linux-2.0, it's in
  Reiter> kernel/sys.c.)  Fortunately, the floating point format used is such
  Reiter> that the numbers aren't far off (IISeeC).
  Reiter> 
  Reiter> Note that I've also submitted a bug report to the author of the
  Reiter> kernel code and suggested that he might consider changing format
  Reiter> again, and also suggested (independent of such a change) a couple
  Reiter> of other changes to acct.h.  Therefore, I suggest that you either
  Reiter> wait until include/linux/acct.h changes again in the 2.1 series
  Reiter> before seriously trying to address the problem (other than perhaps
  Reiter> in the package description and/or testing uname -r during
  Reiter> installation), or else I'll send you a copy of the message I sent
  Reiter> to MvW.

-- 
mailto:edd@debian.org              According to the latest official figures, 
http://rosebud.ml.org/~edd      43% of all statistics are totally worthless.

Send a report that this bug log contains spam.