svn commit: r244154 - head/bin/ps

[Bruce Evans]

On Thu, 13 Dec 2012, Pawel Jakub Dawidek wrote:

> On Thu, Dec 13, 2012 at 08:16:21PM +0200, Konstantin Belousov wrote:
>> On Thu, Dec 13, 2012 at 05:55:41PM +0100, Pawel Jakub Dawidek wrote:
>>> On Thu, Dec 13, 2012 at 06:12:42PM +0200, Konstantin Belousov wrote:
>>>> On Thu, Dec 13, 2012 at 12:12:44PM +0100, Pawel Jakub Dawidek wrote:
>>>>> On Wed, Dec 12, 2012 at 11:06:52PM +0200, Konstantin Belousov wrote:
>>>>> I saw CTLFLAG_TUN on the sysctl and assumed it is read-only...
>>>>> How about defining BSD_PID_MAX in sys/proc.h, which would be visible by
>>>>> userland as well and setting PID_MAX to BSD_PID_MAX?
>>>>>
>>>>> This would also help bsnmpd.
>>>>>
>>>>> 	http://people.freebsd.org/~pjd/patches/PID_MAX.patch
>>>> Do you know why PID_MAX is under _KERNEL ? If there is no real reason,
>>>> it would be better to move it outside kernel-only section. sys/proc.h
>>>> is not in POSIX anyway.

I don't really know, but POSIX says that {PID_MAX} is intentionally left
out of POSIX because pids_t might be cookies in a very large address
space so that there is no useful use of {PID_MAX}.  (POSIX doesn't
say exactly this.  It says "arrays of values of this type [uid_t, gid_t
or pid_t] are unlikely to be fully portable".)

>>> I assumed it will break some ports that may define it themselves.
>>> I wonder if we could do a test ports build to see what's the impact.
>>
>> Sure.
>>
>> On the other hand, sys/proc.h is mostly useless for the application code
>> as it is now. Might be, use
>> #ifndef PID_MAX
>> braces ?

Ugh.  If there is any useful use of {PID_MAX}, then this just breaks
detection of using the wrong value.

> This can be done of course, but it won't help cases where PID_MAX is
> defined after sys/proc.h is included.

Bruce