Re: SQUID_DEFINE_BOOL and broken ifdefs

From: Kinkie <gkinkie_at_gmail.com>
Date: Wed, 7 Jul 2010 16:28:31 +0200

On Wed, Jul 7, 2010 at 4:11 PM, Alex Rousskov
<rousskov_at_measurement-factory.com> wrote:
> On 07/07/2010 07:29 AM, Kinkie wrote:
>> On Tue, Jul 6, 2010 at 11:22 PM, Alex Rousskov
>> <rousskov_at_measurement-factory.com> wrote:
>>> Hello,
>>>
>>>    The kill-parent hack got magically enabled on its own in trunk. The
>>>  problem is the "#ifdef KILL_PARENT_OPT" line in main.cc. Since
>>> KILL_PARENT_OPT is set to "0" when the option is disabled, the quoted
>>> preprocessor condition is always true.
>>>
>>> A simple script found a couple of more potential bugs:
>>>
>>> ./src/main.cc:593:#ifdef KILL_PARENT_OPT
>>> ./include/cache_snmp.h:7:#ifdef SQUID_SNMP
>>
>> I've checked all SQUID_DEFINE_BOOL's in trunk; these appear to be the
>> only cases.
>> I've committed a fix.
>>
>>> ./include/profiling.h:69:#undef USE_XPROF_STATS
>>
>> This is intenional. It is conditional on the CPU/OS being used not
>> supporting profiling infrastructure.
>
> Regardless of the intent, is not that an inconsistency? Should we set
> USE_XPROF_STATS to 0 instead of just undefining it to be consistent with
> what configure.in does?

The effect is the same, but you are right.
Fixing.

-- 
    /kinkie
Received on Wed Jul 07 2010 - 14:28:40 MDT

This archive was generated by hypermail 2.2.0 : Wed Jul 07 2010 - 12:00:40 MDT