Re: ruby and NOT_PAX_MPROTECT_SAFE (was: CVS commit: pkgsrc/lang)

[Joerg Sonnenberger <joerg%bec.de@localhost>]

On Fri, Oct 11, 2019 at 12:53:28PM +0900, Ryo ONODERA wrote:
> Hi,
> 
> Joerg Sonnenberger <joerg%bec.de@localhost> writes:
> 
> > On Thu, Oct 03, 2019 at 04:59:07PM +0900, Ryo ONODERA wrote:
> >> I marked these four .so files with paxctl +m.
> >
> > The attributes are a process property, not a DSO property.
> 
> I see. Thank you.
> 
> >> And I have no idea about Python case, however I think that all ruby-ffi
> >> or py-cffi consumers may not hit the same errors.
> >
> > Python is fine with it. devel/libffi supports closures properly.
> >
> > Can you please make sure that ruby-ffi is not using its own copy of
> > libffi, but a proper dependency?
> 
> It seems that two ffi_c.so files depend on libffi.so properly.

So looking into the source, they implement their own closure management
instead of using the one of libffi. *sigh*

Joerg