Comment # 5 on bug 1169468 from Richard Biener

(In reply to Michael Matz from comment #3)
> (In reply to Richard Biener from comment #2)
> > That would also need a more explicit dependency on python though, no?
> 
> % ldd /usr/bin/gdb
> ...
> libpython3.7m.so.1.0 => /usr/lib64/libpython3.7m.so.1.0 (0x00007efc357de000)
> ...
> 
> It already _has_ very explicit dependencies on python, so this would just
> clarify
> that.  But I'm wondering if this change alone helps with pyenv?
> 
> > Not sure if we should care at all - is pyenv some "supported" feature?
> > I suspect many packages fall foul of its behavior.
> 
> Yeah, but if we can robustify gdb somewhat that's appreciated.  Of course,
> if pyenv uses too forceful means to override the system python at runtime,
> then
> all bets are off (and indeed, other system packages requiring python will
> probably break in similarly mysterious ways).

It smells like one part of it is altering LD_LIBRARY_PATH since reporter
mentions we should set RPATH.  The other part is altering pythons default
search path for sure.

I guess a better(?) way would be if gdb would disable its python support
in a less verbose way when it detects these issues.