Re: Overriding LN_S

[Peter Rosin]

Den 2010-09-02 21:45 skrev Peter Rosin:
> Den 2010-09-02 21:32 skrev Ralf Wildenhues:
>> Hi Peter,
>>
>> * Peter Rosin wrote on Thu, Sep 02, 2010 at 09:18:29PM CEST:
>>> For various reasons I want to override the decision autoconf
>>> makes regarding LN_S. It it set to "ln -s" on my system which
>>> is perfectly fine in every normal case. But in my abnormal case
>>> I want it to be "cp -p". What can I do?
>>
>> I'm rather sure either LN_S or as_ln_s used to be overridable at one
>> point, but doesn't seem to be the case in the current code.  Hmm, I
>> see ac_cv_prog_LN_S in Autoconf from ten years ago, but that code had
>> other problems too.
>>
>> One workaround would be to write a wrapper program called ln which does
>> exit 1.  :-)  Or, if you somehow need links for something else, fail if
>> source and target match conf[1-9][0-9]*.* (regex not glob notation)
>> after disregarding -s, that ought to catch the typical configure tests.
> 
> Yes, that will probably work, but it would be so much easier to be able
> to say .../configure LN_S="cp -p"
> 
>>> I also want this to
>>> propagate to recursive autoconf invocations, so it is not
>>> sufficient to manually edit the generated configure script.
>>
>> You mean recursive configure invocations?
> 
> No, I meant that some libtool tests invoke autoconf, and then configures
> the result, e.g. libtool/tests/recursive.at
> If I hack up my top level configure to go with LN_S="cp -p", that will
> not matter when the testsuite gets to recursive.at, since it will
> reset LN_S to "ln -s". At least I think it will, I haven't actually
> tried...

Aha, it seems it is enough if I invoke the libtool testsuite as

make check-local LN_S="cp -p"

but I will have to remember the LN_S part, which is going to bite me at
some point.

Cheers,
Peter