[Olsr-dev] please revert " force copy the olsrd.conf file during install"

[Hans-Christoph Steiner]

On Jul 4, 2012, at 5:16 PM, Ferry Huberts wrote:

> 
> 
> On 04-07-12 22:23, Hans-Christoph Steiner wrote:
>> 
>> On Jul 4, 2012, at 3:15 PM, Ferry Huberts wrote:
>> 
>>> 
>>> 
>>> On 04-07-12 11:51, Vasilis Tsiligiannis wrote:
>>>> Στις Τετ 04 Ιουλ 2012 09:07:51 ZioPRoTo έγραψε:
>>>>> the upstream make install will install olsrd and plugins at
>>>>> once. Instead OpenWRT packages each plugin in a separate ipkg,
>>>>> so I think it is impossible to use the upstream make install.
>>>> 
>>>> Hi Saverio,
>>>> 
>>>> This is not a problem. The packaging of the files is performed at
>>>> a later stage. Just "fish" the files from PKG_INSTALL_DIR instead
>>>> of PKG_BUILD_DIR.
>>>> 
>>>> One problem with using the upstream 'make install{,_libs}' is the
>>>> hardcoded use of host 'strip' for striping the object files. It
>>>> should be wrapped into a variable which will allow it to be
>>>> overriden in the OpenWrt makefile.
>>> 
>>> Please enlighten me,
>>> 
>>> Why would you not want to strip when building without debug?
>> 
>> Because then you don't need to have a separate debug mode.  If the
>> software can be built and run in place, i.e. run ./olsrd in the
>> source tree after its built, then you can debug it without
>> installing.  I do this all the time, setting LD_LIBRARY_PATH or using
>> symlinks to load plugins.
>> 
> 
> yes. I do that too.
> However, I don't debug a no-debug build and I bet you don't either...
> 
> Distribution / release builds should use 'make DEBUG=0 install'.
> 
> 
> So, any other answer?


In the vast majority of build systems, "make install" already means "build and install a stripped version".  So DEBUG=0 is a custom thing that needs to be documented, learned, remembered.  A set of commands is like a library in our brains.  Using that "library" means the documenting, learning, and remembering part doesn't need to be implemented since its already there.

.hc