[PATCH] Get library installation path for target architecture

[Bernhard Fischer]

On Fri, Mar 24, 2006 at 02:41:43PM -0500, Mike Frysinger wrote:
>On Friday 24 March 2006 14:37, Rob Landley wrote:
>> On Thursday 23 March 2006 10:48 pm, Mike Frysinger wrote:
>> > On Thursday 23 March 2006 22:30, Rob Landley wrote:
>> > > The situation I'm actually wondering about is a cross-compile where
>> > > during the build the libraries live
>> > > at /opt/totally/ludicrous/deep/path/to/cross/compiler and then on the
>> > > target system they live in "/lib".  Will that do the right thing?
>> >
>> > if you were to take a non-multilib cross-compiler (say targeting amd64)
>> > and deploy it on a multilib host (say amd64 with x86 compat), then
>> > chances are it'd be in the wrong dir
>>
>> So if you use a cross-compiler that doesn't match the host you're deploying
>> on, you have a problem.
>>
>> This is news?
>
>there is absolutely nothing wrong with the scenario i stated above ... it is 
>perfectly valid to have a multilib host and to have a non-mulitlib 
>cross-compiler help it out on another machine

The only thing that really works for the lib path is a config option.
Some may not care, but a about the half of the few commercial compilers
i have access to don't have a gcc-driver thus just cannot cope with any
-print-this-or-that option gcc may support to find the correct dir to
install that stuff to.

Rob, that -print-file-name=libc.so which is in the tree now breaks real
setups, fwiw.