Re: PING: [PATCH] Add --enable-static-pie to build static PIE

[Joseph Myers <joseph at codesourcery dot com>]

On Wed, 27 Sep 2017, H.J. Lu wrote:

> > I think the most useful test for this purpose (of assessing architecture
> > support) is adding --enable-static-pie alongside --enable-profile in
> > build-many-glibcs.py and seeing how the build goes.
> 
> I will do that.

Thanks.  To be clear, I don't think such a build-many-glibcs.py change 
would be appropriate for mainline (given a version of --enable-static-pie 
that changes how libc.a is built rather than enabling building additional 
static-PIE libraries) - although adding one or two additional static-PIE 
configurations to build-many-glibcs.py might be - but for testing purposes 
it's very useful to understand how widely this option works (or at least 
builds) for other architectures.

> > To me, both the install.texi and NEWS changes read as saying that programs
> > such as ldconfig are built as PIEs.  That's a separate matter from
> 
> All static executables, including tests,  within glibc are built as
> static PIE when
>  --enable-static-pie is used.  That is how I noticed that
> --eh-frame-hdr is needed
> for static PIE.

My point is that NEWS and the install.texi text are for users of glibc, 
and I think it's more significant for users that they can build *their 
own* programs as static PIEs, than that ldconfig and sln are static PIEs.  
So the NEWS and install.texi text should talk about the effects of the 
option on what users can do with the resulting glibc.

-- 
Joseph S. Myers
joseph@codesourcery.com