diff mbox series

[3/4] package/bind: don't enable server by default

Message ID 20190423212406.13909-3-fontaine.fabrice@gmail.com
State Superseded
Headers show
Series [1/4] package/bind: enable static build | expand

Commit Message

Fabrice Fontaine April 23, 2019, 9:24 p.m. UTC
Don't enable server by default as dhcp will now select bind to use the
bind libraries

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
---
 package/bind/Config.in | 1 -
 1 file changed, 1 deletion(-)

Comments

Thomas Petazzoni April 24, 2019, 8:36 p.m. UTC | #1
Hello,

On Tue, 23 Apr 2019 23:24:05 +0200
Fabrice Fontaine <fontaine.fabrice@gmail.com> wrote:

> Don't enable server by default as dhcp will now select bind to use the
> bind libraries
> 
> Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
> ---
>  package/bind/Config.in | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/package/bind/Config.in b/package/bind/Config.in
> index 60546f691b..6d4ee2e6d7 100644
> --- a/package/bind/Config.in
> +++ b/package/bind/Config.in
> @@ -28,7 +28,6 @@ if BR2_PACKAGE_BIND
>  
>  config BR2_PACKAGE_BIND_SERVER
>  	bool "Install server components"
> -	default y

This will break existing defconfigs which did not explicitly enable
BR2_PACKAGE_BIND_SERVER because it was implicitly enabled when
BR2_PACKAGE_BIND=y. But I don't see a good solution to avoid that. We
certainly don't want BR2_PACKAGE_DHCP=y to install the full BIND server.

So to me, it's an acceptable trade-off. I've Cc'ed Arnout, Peter and
Yann to see what they think.

Best regards,

Thomas
Arnout Vandecappelle April 24, 2019, 8:56 p.m. UTC | #2
On 24/04/2019 22:36, Thomas Petazzoni wrote:
> Hello,
> 
> On Tue, 23 Apr 2019 23:24:05 +0200
> Fabrice Fontaine <fontaine.fabrice@gmail.com> wrote:
> 
>> Don't enable server by default as dhcp will now select bind to use the
>> bind libraries
>>
>> Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
>> ---
>>  package/bind/Config.in | 1 -
>>  1 file changed, 1 deletion(-)
>>
>> diff --git a/package/bind/Config.in b/package/bind/Config.in
>> index 60546f691b..6d4ee2e6d7 100644
>> --- a/package/bind/Config.in
>> +++ b/package/bind/Config.in
>> @@ -28,7 +28,6 @@ if BR2_PACKAGE_BIND
>>  
>>  config BR2_PACKAGE_BIND_SERVER
>>  	bool "Install server components"
>> -	default y
> 
> This will break existing defconfigs which did not explicitly enable
> BR2_PACKAGE_BIND_SERVER because it was implicitly enabled when
> BR2_PACKAGE_BIND=y. But I don't see a good solution to avoid that. We
> certainly don't want BR2_PACKAGE_DHCP=y to install the full BIND server.
> 
> So to me, it's an acceptable trade-off. I've Cc'ed Arnout, Peter and
> Yann to see what they think.

 Although not documented anywhere, in my opinion a defconfig is not supposed to
be stable over upgrades. When upgrading, you should use a full config.

 I do the same for the kernel as well.

 I also always do a diffconfig between old and new and try to understand the
differences.

 Regards,
 Arnout
Yann E. MORIN April 25, 2019, 3:56 p.m. UTC | #3
Arnout, All,

On 2019-04-24 22:56 +0200, Arnout Vandecappelle spake thusly:
> On 24/04/2019 22:36, Thomas Petazzoni wrote:
> > On Tue, 23 Apr 2019 23:24:05 +0200
> > Fabrice Fontaine <fontaine.fabrice@gmail.com> wrote:
> > 
> >> Don't enable server by default as dhcp will now select bind to use the
> >> bind libraries
> >>
> >> Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
> >> ---
> >>  package/bind/Config.in | 1 -
> >>  1 file changed, 1 deletion(-)
> >>
> >> diff --git a/package/bind/Config.in b/package/bind/Config.in
> >> index 60546f691b..6d4ee2e6d7 100644
> >> --- a/package/bind/Config.in
> >> +++ b/package/bind/Config.in
> >> @@ -28,7 +28,6 @@ if BR2_PACKAGE_BIND
> >>  
> >>  config BR2_PACKAGE_BIND_SERVER
> >>  	bool "Install server components"
> >> -	default y
> > 
> > This will break existing defconfigs which did not explicitly enable
> > BR2_PACKAGE_BIND_SERVER because it was implicitly enabled when
> > BR2_PACKAGE_BIND=y. But I don't see a good solution to avoid that. We
> > certainly don't want BR2_PACKAGE_DHCP=y to install the full BIND server.
> > 
> > So to me, it's an acceptable trade-off. I've Cc'ed Arnout, Peter and
> > Yann to see what they think.
> 
>  Although not documented anywhere, in my opinion a defconfig is not supposed to
> be stable over upgrades. When upgrading, you should use a full config.

Agreed. The process I follow to upgrade buildroot if (roughly):

    $ make my_foo_defconfig
    $ git pull
    $ make olddefconfig
    $ diff -du .config.old .config
    $ make savedefconfig
    $ git diff   # shows delta in my_foo_defconfig

>  I also always do a diffconfig between old and new and try to understand the
> differences.

This sounds all too familiar, too. ;-)

Regards,
Yann E. MORIN.
diff mbox series

Patch

diff --git a/package/bind/Config.in b/package/bind/Config.in
index 60546f691b..6d4ee2e6d7 100644
--- a/package/bind/Config.in
+++ b/package/bind/Config.in
@@ -28,7 +28,6 @@  if BR2_PACKAGE_BIND
 
 config BR2_PACKAGE_BIND_SERVER
 	bool "Install server components"
-	default y
 	help
 	  Install server components (named, lwresd)