[PATCH] Get library installation path for target architecture

Rob Landley rob at landley.net
Fri Mar 24 03:18:04 UTC 2006


On Thursday 23 March 2006 6:55 pm, Mike Frysinger wrote:
> On Thursday 23 March 2006 17:12, Pavel Roskin wrote:
> > Since Rules.mak exports CC, it could be used to find the library path.
> > "gcc -print-file-name=libc.so" would print path to libc.so, and the last
> > part of path before libc.so is what we need.  For safety reasons, only
> > paths starting with "lib" are matched.  In case the compiler fails to
> > produce something we recognize, /lib is the default value.
>
> the current code is a hack ... the proposed solution is a hack ...

A less sucky hack, but sure.

I merged the patch already.  That's not necessarily the final word, but I'd 
like to point out it was submitted by someone experiencing the actual 
problem, and fixed it for them without (as far as I know) breaking the setup 
for anyone else whose setup wasn't already broken.  I consider that an 
improvement.

> i say we  just make it a config option, default it to 'lib', and then make
> the user override it if 'lib' rubs them the wrong way
>
> there is no consensus yet (if there ever will be) as to what the "correct"
> location is on any system, whether it be multilib or not

I liked /lib.  I knew what /lib meant.  The x86-64 guys who decided that they 
needed /lib64 instead of moving the _32_bit_ stuff to a new location...  That 
was sad.

We already have a busybox installation prefix config entry thing under 
"installation options".  It would be nice if we could make intelligent use of 
this and append some stable path (usr/lib, perhaps?) onto it.

I dislike punting to the user when it makes the common case of configuring 
busybox more work for everybody.

> -mike

Rob
-- 
Never bet against the cheap plastic solution.



More information about the busybox mailing list