limits in /etc/passwd, and maybe a bug in processing /etc/limits? :)

[Jim B <jamesb-debian@alongtheway.com>]

OK.  As a continuation of my previous ramblings on resource limits, I'm
running into two more similar issues on my slink machine.

According to /etc/login.defs, I should be able to employ resource limits
by editing users' passwd entries.  I have "QUOTAS_ENAB" in login.defs:

# Enable setting of ulimit, umask, and niceness from passwd gecos field.
#
QUOTAS_ENAB             yes


If I look at man 5 passwd, I see the following:


       The  comment  field  is  used by various system utilities,
       such as finger(1).  Three additional values may be present
       in the comment field.  They are

            pri= - set initial value of nice
            umask= - set initial value of umask
            ulimit= - set initial value of ulimit

       These  fields  are  separated from each other and from any
       other comment field by a comma.


I tried to set the umask to 022 this way with a test account, and I can't
get it to do anything at all.

I have tried adding "extra" comment entries by adding commas in
/etc/passwd, and I've also tried using the pre-existing comment
entries.  None of it works.  I end up with the default umask of 002 no
matter what... and yes I have commented-out the umask field in
/etc/profile, and there is none in the test user's .bash_profile,
.profile, and .bashrc.  :)

Anyone know the right way to do it?


My second problem... well, it looks like it may be a bug.  Note the
following text in /etc/limits:

# Valid flags are:
# A: max address space (KB)
# C: max core file size (KB)
# D: max data size (KB)

... and so on.


But any time I use the "A" limit, the whole line becomes useless.  See the
following in man 5 limits:

"A invalid limits string will be rejected (not considered) by the login
program."

If I take out the "A" limit, the rest of the line functions again.  So
there seems to be some kind of problem reading or enforcing this limit.


So a line like this:

* L2 D12288 M32768 R2048 S2048 U64 N256 F16384 T60 C0

works fine.


But one like this:

* A32768 L2 D12288 M32768 R2048 S2048 U64 N256 F16384 T60 C0

breaks the whole line and NO limits are enforced.


Is this a bug or am I doing something wrong (again)?  :)