[OpenWrt-Devel] [PATCH v2 2/2] dnsmasq: fix dependency problems of the dnsmasq-full variant.

[Yousong Zhou]

On 2 January 2015 at 02:05, Frank Schäfer <fschaefer.oss at googlemail.com> wrote:
>
> Hi Yousong,
>
> Am 31.12.2014 um 11:48 schrieb Yousong Zhou:
>>
>> ...
>>> tmp/.config-package.in:93:error: recursive dependency detected!
>>> tmp/.config-package.in:93:      symbol PACKAGE_dnsmasq-full depends on
>>> PACKAGE_dnsmasq_full_ipset
>>> tmp/.config-package.in:124:     symbol PACKAGE_dnsmasq_full_ipset
>>> depends on PACKAGE_dnsmasq-full
>>>
>>> If you make a menuconfig, the IPSET build option also doesn't show up
>>> unless you have selected DNSSEC...
>> There are 2 patches in this series, and the first one is needed to
>> avoid the recursive dependency problem [1] from showing up (it's a
>> necessary workaround for now).
>>
>> [1] https://lists.openwrt.org/pipermail/openwrt-devel/2014-December/030211.html
>
> Ah... ok. Didn't notice that it touches ipset.
> Works fine, sorry for the noise.
>
>
>>>> diff --git a/package/network/services/dnsmasq/patches/200-let-makefile-be-aware-of-config-changes.patch b/package/network/services/dnsmasq/patches/200-let-makefile-be-aware-of-config-changes.patch
>>>> new file mode 100644
>>>> index 0000000..663f0e9
>>>> --- /dev/null
>>>> +++ b/package/network/services/dnsmasq/patches/200-let-makefile-be-aware-of-config-changes.patch
>>>> @@ -0,0 +1,42 @@
>>>> +Index: dnsmasq-2.72/Makefile
>>>> +===================================================================
>>>> +--- dnsmasq-2.72.orig/Makefile       2014-12-30 19:46:10.484921940 +0800
>>>> ++++ dnsmasq-2.72/Makefile    2014-12-30 19:56:39.712926794 +0800
>>>> +@@ -74,6 +74,10 @@
>> <snip>
>>
>>> Urgh... is it really necessary to patch dnsmasqs native Makefile ?
>>> Can you give us a practical and reproducable example that shows the
>>> problem you are trying to solve here ?
>> Patching Makefile of dnsmasq itself is necessary for the build system
>> of OpenWrt and Makefile of dnsmasq to pick up a possibly updated
>> config of dnsmasq-full variant.  Otherwise rebuild of dnsmasq-ful may
>> not happen as expected when users enable or disable a feature with
>> e.g. make menuconfig.
>
> Ok, I can reproduce it now...
> Is this patch suitable for upstream inclusion ?
>

An fixed up version of this patch is already in upstream dnsmasq now.
But another one has to be applied to fix possible race condition from
happening when doing parallel build.  I am truly sorry but it seems
like there are just too many cases to be tested out in each iteration.

[1] http://lists.thekelleys.org.uk/pipermail/dnsmasq-discuss/2015q1/009068.html

Regards.

               yousong
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel