[cfe-commits] Fix search path for clang on latest DragonFly releases

[John Marino]

On 11/26/2012 14:55, Rafael Espíndola wrote:
>> Yes, it is passing multiple -L's to the linker.  Is that in itself a
>> problem?  Doesn't the linker check in the -L order and stop when the sought
>> library is found?
>
> It can get you very interesting bugs if the contents of the
> directories are different.

I can confirm that /usr/lib/gcc47 has every library that /usr/lib/gcc44 
has plus additional ones.  I understand mixing libs would be a mess 
waiting to happen but it should not be possible for mixing to occur.


>
>> As far as testing, this patch has been applied to the pkgsrc clang 3.1
>> package.  I know it works for the current release featuring gcc 4.7, and
>> verifying gcc 4.4 should be a simple matter of checking an earlier DragonFly
>> release that doesn't have gcc 4.7 available, so I can do that.
>>
>> I was just trying to keep the patch simple.
>
> It *needs* a test anyway. The code in the driver needs refactoring, so
> it is double important that we test it. The test should go to
> clang/test/Driver/.


I'm not quite clear on this point.  Are you saying that even if I had 
not proposed a patch, the driver would need refactoring?  And because we 
are considering a patch, now is the time to do this?  (I need to review 
the driver code which I don't have ATM to see what code segments need to 
be factored out, that's also not clear).

What should be the objective of this test?  to see if it can produce an 
executable?

John