I am using solaris 8 and the project I am on is using /usr/ucb/ps - gwwux to search for processes that it is supposed to be checking on. The trouble is, every now and then ps core dumps! Here is a stack trace:
On Mon, 21 Jan 2008, apm35@student.open.ac.uk wrote:
I am using solaris 8 and the project I am on is using /usr/ucb/ps -
gwwux to search for processes that it is supposed to be checking on.
The trouble is, every now and then ps core dumps! Here is a stack
trace:
Not sure what the problem you're seeing is, but I'd steer clear of stuff in /usr/ucb if I were you. Those commands and interfaces are historical and you should be using their more modern equivelents in /usr/bin. Does doing that fix your problem?
On 21 Jan, 16:32, Rich Teer <rich.t...@rite-group.com> wrote:
On Mon, 21 Jan 2008, ap...@student.open.ac.uk wrote:
I am using solaris 8 and the project I am on is using /usr/ucb/ps -
gwwux to search for processes that it is supposed to be checking on.
The trouble is, every now and then ps core dumps! Here is a stack
trace:
Not sure what the problem you're seeing is, but I'd steer clear of
stuff in /usr/ucb if I were you. Those commands and interfaces are
historical and you should be using their more modern equivelents in
/usr/bin. Does doing that fix your problem?
No. The reason I am using the BSD ps is because it gives me access to the complete command line. The stuff I am grep'ing for is an invocation of the JVM. Such command lines are extremely long due to the classpath argument and numerous java property settings. System V ps returns truncated output in these cases.
On 21 Jan, 17:26, Doug McIntyre <mer...@geeks.org> wrote:
ap...@student.open.ac.uk writes:
I am using solaris 8 and the project I am on is using /usr/ucb/ps -
gwwux to search for processes that it is supposed to be checking on.
The trouble is, every now and then ps core dumps!
I've run /usr/ucb/ps on Solaris 8 probably thousands of times in my
lifetime on dozens if not hundreds of machines. Its never core dumped
on me before.
Lucky you.
BSD ps is just one of those things I never could get
away from using. Especially now with being on more FreeBSD and Darwin
systems than Solaris.
This is not a matter of personal taste for me. My taste happens to be for System V but in this case the BSD command is forced upon me. It is the only way I know of to get the complete command. System V ps tends to truncate long lines. And the lines I am after are long because they are JVM invocations with long classpath settings and numerous java properties.
I notice its calling getenv from localtime_r. So just in case its
relevant I will say that the value for TZ is 'GB-Eire'.
It could be the environment.
Er, I suppose it could. Can you be more specific? What did you have in mind?
On 22 Jan, 12:01, Henry Townsend <henry.towns...@not.here> wrote:
ap...@student.open.ac.uk wrote:
This is not a matter of personal taste for me. My taste happens to be
for System V but in this case the BSD command is forced upon me. It is
the only way I know of to get the complete command. System V ps tends
to truncate long lines. And the lines I am after are long because they
are JVM invocations with long classpath settings and numerous java
properties.
man pargs
Well, I've learnt something new. Thanks for that. But it will be a bit of a pain to use coz I will have to use ps to get the pids then invoke pargs in a loop getting the args one pid at a time. But still doable and avoids BSD ps. Cheers!
On 22 Jan, 16:45, Doug McIntyre <mer...@geeks.org> wrote:
I am using solaris 8 and the project I am on is using /usr/ucb/ps -
gwwux to search for processes that it is supposed to be checking on.
The trouble is, every now and then ps core dumps!
I notice its calling getenv from localtime_r. So just in case its
relevant I will say that the value for TZ is 'GB-Eire'.
There could be some weird interaction with the older not-so-maintained
program dealing with a charset other than C/POSIX.
Ie. LC_* for me has typically been set to C or en_US.ISO8859-1.
The routines dealing with parsing the data and putting out localized
locale messages might be buggy interacting with the old /usr/ucb/pc code.
Er, I suppose there could be a bug that is decades old that only bites users that are not based in the USA. A bug in one of the most common commands, ps, undetected for all this time. But it seems unlikely to me. I do not rule out the possibility of bugs, which I why I started this thread. But I was hoping for something more specific than "there might be a bug in ps".
Thomas Schulz 23 January 2008 22:32:40 [ permanent link ]
In article <9a5db800-810d-4f64-8c96-2ba458c103ee@e10g2000prf.googlegroups.com>, <apm35@student.open.ac.uk> wrote:
On 21 Jan, 17:26, Doug McIntyre <mer...@geeks.org> wrote:
ap...@student.open.ac.uk writes:
I am using solaris 8 and the project I am on is using /usr/ucb/ps -
gwwux to search for processes that it is supposed to be checking on.
The trouble is, every now and then ps core dumps!
I've run /usr/ucb/ps on Solaris 8 probably thousands of times in my
lifetime on dozens if not hundreds of machines. Its never core dumped
on me before.
Lucky you.
BSD ps is just one of those things I never could get
away from using. Especially now with being on more FreeBSD and Darwin
systems than Solaris.
This is not a matter of personal taste for me. My taste happens to be
for System V but in this case the BSD command is forced upon me. It is
the only way I know of to get the complete command. System V ps tends
to truncate long lines. And the lines I am after are long because they
are JVM invocations with long classpath settings and numerous java
properties.
I notice its calling getenv from localtime_r. So just in case its
relevant I will say that the value for TZ is 'GB-Eire'.
It could be the environment.
Er, I suppose it could. Can you be more specific? What did you have in
mind?
If you are running most any shell but csh you should be able to type: LANG=C /usr/ucb/ps your-arguments-here If you are running csh then you will have to use 'setenv LANG C'
On 23 Jan, 14:33, Michael Tosch <eed...@NO.eed.SPAM.ericsson.PLS.se> wrote:
I am using solaris 8 and the project I am on is using /usr/ucb/ps -
gwwux to search for processes that it is supposed to be checking on.
The trouble is, every now and then ps core dumps!
You might want to apply the ps patch 109023-07.
Michael Tosch @ hp : com
That patch does not mention a fix for random core dumping. Here is the list of problems the patch fixes:
--- 6536923 T109023-06 and T109024-06 are delivering ps without SUID bit (from 109023-06) 6491490 /usr/ucb/ps is filling /tmp with files ps.chars (from 109023-05) 4798073 /usr/ucb/ps -e should display environment variables only for user-owned processes 5045071 /usr/ucb/ps gxuaww only displays caller's processes (from 109023-04) 4988595 defunct processes show the output as "0" instead of "-" (from 109023-03) 4519180 ps shows date of Jan_01 for defunct processes when using -o stime 476669 ps utility uses the FORBIDDEN 'cftime' instead of strftime function 4785109 ps core dump with -o etime (from 109023-02) 4559012 /usr/ucb/ps hangs after reboot until one is killed creating / tmp/ps_data ----
