Re: [PATCH] libtool ignores unresolved symbols despite -no-undefined flag

[Torsten Landschoff]

Hi Ralf,

On Sun, Nov 15, 2009 at 02:36:41PM +0100, Ralf Wildenhues wrote:

> * Torsten Landschoff wrote on Sun, Nov 15, 2009 at 04:16:34AM CET:
> > I think, GNU libtool should indead report an error if building with
> > -no-undefined even on systems where undefined symbols are allowed. That
> 
> Yes, this is a good idea.  The problem with this is that it frequently
> happens that some of the system libraries like glibc have undefined
> symbols themselves, on purpose, subverting this nice feature.  (The
> symbols are then somehow satisfied at runtime.)
> 
> At least that's how it used to be last time we tried this.  I haven't
> checked this in a long time.  It may be that this only mattered for
> C++ or so.  But anyway changing this would require somebody to dig
> this out and verify that the issue is gone for the systems/libc
> versions where we enable it.

I have to admit that I tested it only on Debian unstable. Maybe other systems
indeed show this behaviour. But in that case, that should be documented I'd
say. Interestingly, libtool seems to use -z defs on some systems.

> FWIW, in your patch series, I'd squash patches 1,2,5, and patches 3,4.
> A Fortran test would be good.

I can merge the patches as requested. So far, I do not speak Fortran. My
bookshelf contains some Fortran book so, so I might try to write such a
test.

> Please also note that in order to accept nontrivial patches, we need
> copyright assignment papers from you (details off-list).  If that won't
> work out for you, then it is better if you don't post patches, but only
> explain needed changes, so that there is no doubt someone else
> reimplementing them causes copyright issues.

I don't have any problem with assigning copyright (or copyleft) to the FSF.

Greetings, Torsten