Re: [FYI] Cleanup of linux pass_all

[Howard Chu]

Bob Friesenhahn wrote:
> On Wed, 15 Sep 2004, Alexandre Oliva wrote:
>
> > > if there was only a good way to test for PIC ....
> >
> >
> > The only reasonable way I can think of is to get the linker to try and
> > link stuff.  If it fails, we print a warning message and proceed to
> > building a static-only version of the library.  And let's just hope
> > modern linkers won't silently create corrupt binaries.
>
>
> I expect that assumption takes a lot of finger-crossing.  Some problems 
> will be found at run-time by ld.so and some will be found via gdb in the 
> resulting core dump. :-)

All of which should be fairly obvious as soon as the person compiling 
the code tries to execute the result.

When a developer writes code assuming the existence of dynamic 
libraries, and the platform that's being used doesn't provide equivalent 
support, it's pretty likely that there are other non-portable 
assumptions in the code itself. I.e., even after libtool bends over 
backwards to make the program link, it's likely that there are other 
issues in the code that will probably overshadow the question of "does 
it link?" Spending a ton of effort on these questions of linking isn't 
going to make anybody's life easier.
--
  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc
  Symas: Premier OpenSource Development and Support