>>>>> "Earnie" == Earnie Boyd <[EMAIL PROTECTED]> writes:
>> I didn't know stow would rely on this. But still, it seems
>> dangerous to me: why don't you ./configure
>> --prefix=/usr/stow/mypackage.mj.mn.p in order to be sure that the
>> hard coded paths are right?
>>
Earnie> No. The hard coded paths need to reflect the configured
Earnie> prefix which reflects where stow will symlink the files. The
Earnie> stow documentation warns of badly configured Makefiles that
Earnie> will rebuild when specifying `make install prefix=/foo'; but
Earnie> as the standards documentation you quoted stated that isn't
Earnie> supposed to happen.
Yep, this is a great news. I believe quite many of us here thought
this was not what was expected (hence my messages: the other
interpretation seemed and is insane).
>> | configure (prefix=/usr/local) | make prefix=/usr (user Foo always
>> does it this way) | make install | | I may want to test an
>> installation and want to build with a different prefix | than what
>> it was configured with.
>>
>> But again, I know the goal. I just say that it is difficult to
>> make it right, so you cannot expect the package to behave
>> correctly.
>>
>> The reason why you can't depend upon this is that somewhere, IMHO,
>> it is just wrong to expect make prefix=/foo to work properly.
>> Specifying prefix etc. is a job for configure, not make.
Earnie> I understand your point but you must be flexable on this. I
Earnie> do expect that the prefix is changeable at the time of make.
Earnie> My guess is that others do also.
Yep, but really if you want to understand what I meant, you'll have to
reread with `make install prefix=/foo == expecting the package to work
with prefix=/foo at runtime hence, if there are hard coded paths,
recompile'.
