Re: [PATCH 1/3] cmdline_parse: Also pass bool_assert to OPT_CUSTOM so that parse_bool can be used correctly.

[Don Slutz <dslutz@verizon.com>]

On 07/28/14 12:23, Andrew Cooper wrote:
> On 28/07/14 17:18, Don Slutz wrote:
>>
>> On 07/28/14 12:08, Andrew Cooper wrote:
>>> On 28/07/14 16:59, Don Slutz wrote:
>>>> Based on:
>>>>
>>>> commit 2fcb51bc7a4fcb7534265d7bb155c6ddf03952b8
>>>> Author:kfraser@localhost.localdomain  <kfraser@localhost.localdomain>
>>>> Date:   Mon Jul 9 14:29:53 2007 +0100
>>>>
>>>>      Also allow boolean cmdline params to be inverted in two other ways.
>>>>
>>>>      Now, a standard boolean (e.g., watchdog) can be inverted in three
>>>>      ways:
>>>>       1. no-watchdog
>>>>       2. watchdog=no
>>>>       3. watchdog=off
>>>>
>>>>      Stacked inversions cancel each other: no-watchdog=no turns on the
>>>>      watchdog.
>>>>
>>>> bool_assert is needed to do stacked inversions with parse_bool.
>>>>
>>>> Signed-off-by: Don Slutz<dslutz@verizon.com>
>>>
>>> Why do we need to care about stacked inversions?
>>>
>>> From xen-command-line.markdown:
>>>
>>> "Explicitly specifying any value other than those listed above is 
>>> undefined, as is stacking a |no-| prefix with an explicit value."


Ah, I see that:


commit 6328e728f6d6589a10d7e9f97a47f566f63743c2
Author: Andrew Cooper <Andrew.Cooper3@citrix.com>
Date:   Mon Sep 3 11:22:00 2012 +0100

     docs/command line: Clarify the behavior with invalid input.

     Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
     Acked-by: Jan Beulich <jbeulich@suse.de>
     Committed-by: Ian Campbell <ian.campbell@citrix.com>

was when this changed.

>> The command line:
>>
>> no-console_timestamps dom0_mem=3G loglvl=all guest_loglvl=all 
>> com1=9600,8n1 console=com1 crashkernel=256M@256M no-console_timestamps=no
>>
>> in 4.4 and before ends up with date style console_timestamps. In xen 
>> 4.5 you get no console_timestamps.
>>
>> The simpler case of no-console_timestamps no longer works also.
>>
>>    -Don Slutz
>
> Ok, so my "compat for older console_timestamp boolparam" needs 
> improving a bit.
>

Yes, that is what patch #2 does.


> However, "no-console_timestamps=no" is undefined, and we make no 
> guarentees as to what happens if you give repeated (different) command 
> line options to Xen.  Both of these are user/admin error and should 
> not be worked around in Xen code.
>
> ~Andrew

Ok,  I was not aware of this change in "defined" behavior.  I will 
re-work this
to include Jan's 'pass "no" if "no-" removed.'

    -Don Slutz