cvs commit: src/usr.bin/top machine.c

[Bruce Evans]

On Sun, 15 May 2005, David O'Brien wrote:

> On Mon, May 16, 2005 at 03:40:59AM +0300, Giorgos Keramidas wrote:
>> On 2005-05-15 17:18, David O'Brien <obrien at FreeBSD.org> wrote:
>>>
>>> Please back machine.c back to revsion 1.69, ASAP.
>>> The current top is redicious:
> ..snip..
>> Your USERNAME column is too wide (12 columns).  Do you happen to have
>> very long names?
>
> The dovecot port adds a the 'dovecot-auth' user.
> The top(1) code supports 13 character usernames on SMP systems and 15 on
> UP systems.  Your rev 1.70-1.71 changes failed to take that into account.

Another bug in top is that it expands the USERNAME column to have enough
space for the longest user name (clipped to 13 or 15 characters), even
when the user(s) with the longest name have never run any processes.

Another bug in at last old versions of top (machine.c 1.51) is that this
support is broken if there is actually a user name of maximal length, due
to essentially the same bug that brought this thread back to life: the
header is not properly truncated; for a 15-character USERNAME column it
gives a header length of 80, and on 80-column terminals with auto-wrap
this causes the same horrible line wrap is for header lengths > 80
(but slightly uglier because the extra line doesn't even have anything on
it):

%%%
18 process 15391 running, 17 slee 0.03,  0.01,  0.00    up 0+00:28:54  14:37:01
CPU states:     % user,     % nice,     % system,     % interrupt,     % idle
Mem: 7516K A 0.0e, 17M I 0.0, 20M Wi 0.0 36K Cache 0.0M Buf, 957M F 100
Swap:  64

   PID USERNAME        PRI NICE   SIZE    RES STATE    TIME   WCPU    CPU COMMAND

     1 root              8    0   716K   368K wait     0:01  0.00%  0.00% init
   551 bde               8    0  3200K  3072K wait     0:00  0.00%  0.00% bash
...
%%%

The extra line causes the screen to scroll, and this messes up the lines
before the header.

(I tried adding a 16-character user name to see if utilities handle the
resulting non-strings in struct utmp.  They do.)

>> If that is true, would it be ok to wait 1-2 days until I have a
>> machine.c version that prints USERNAME using only 8 columns instead of
>> using arbitrarily long columns, like now?
>
> I don't know why to wait 1-2 days.  I made the change in 1 minute.  Or is
> there debate if that change will be acceptable?  If there isn't a debate
> about it, let me know and I'll commit this diff.

It is easy to add more bugs in 1 minute.

Bruce