This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: testroot.pristine creation falls over copying dynamic linker
- From: Joseph Myers <joseph at codesourcery dot com>
- To: DJ Delorie <dj at redhat dot com>
- Cc: <carlos at redhat dot com>, <libc-alpha at sourceware dot org>
- Date: Wed, 7 Aug 2019 21:48:12 +0000
- Subject: Re: testroot.pristine creation falls over copying dynamic linker
- Ironport-sdr: 1Dz7L3ZYGe3IIQFPybXxQD81t48ILFXEuuZyEW4PYXrtTuJCimAw7kkATh23jdKiYHZeTbtPq+ 9EiKIR2mVpdoYAv3FhSBtdC9+tsZXSXkzgRRLOLtb/f0Be5/WtdK4I5/PqFz6d75muJNoztSGU ctbioOORhSxag33BC4qyPNx/FlmJh7IT7CAjyZEiaC5JXYv2U1nqdABJXhkyc8GjbsuJTLnJaF DlJ1KQsgTOLKPgy1a4g/WyrtC7esx0S+XIcpF7tP3XNUycMLsHdvu2hHEXWoXgq8QiUzW9VXu9 qBg=
- Ironport-sdr: JAHs6SnAnCABikxHzb6yR2eFkVCNYalaQGrxiCTUYoJnw1WxOetbwMnkV5TgUQ4PoWpIunvEPk +VGktmEWR7p/gWGlgyP11IzMVCQNWblRA4pvogJiGkngPOSYmRPyZodAPpo/4Tkr5iO39TX1IW mKl5oF9beNulH+SLVSW9furX6L01y6+1MO+HLrIL7l97loMc3aKD6zvKxsdvYJ66Ojmh7lELqO iTemGq0gCIJ08vgANM/izrctlaGr4VfK26I9Isrimaw+BPmdflNTZMuUKAFPKBRNraFhJxxWyv ClA=
- References: <xntvb1fxgn.fsf@greed.delorie.com> <alpine.DEB.2.21.1908011646270.31047@digraph.polyomino.org.uk>
On Thu, 1 Aug 2019, Joseph Myers wrote:
> On Wed, 31 Jul 2019, DJ Delorie wrote:
>
> > The case that would fail would be if ldd reported something like this:
> >
> > /a/somelib.so => /b/somelib.so
> >
> > and the dynamic linker expected to find it in /a/
>
> As I understand it, /a/ here is always the canonical path (specified as
> ELF interpreter), which may not actually exist, while /b/ is the path
> actually present outside the testroot, so the only correct one for copying
> from. The glibc shared libraries copied at this stage shouldn't be very
> relevant anyway, only other copied libraries (e.g. libgcc_s) - the glibc
> shared libraries installed by the '$(MAKE) install
> DESTDIR=$(objpfx)testroot.pristine' ought to be the ones that matter.
>
> (It would, however, be desirable for such a change to be tested both with
> and without --enable-hardcoded-paths-in-tests, in addition to any other
> variants that might be relevant for testing.)
I have now done a MIPS test of this patch in the
--enable-hardcoded-path-in-tests case and confirmed it also works in that
case as well - of course in that case it copies libraries to paths under
<builddir>/testroot.pristine/<builddir> so that execution within the
testroot can find the libraries at the hardcoded paths.
--
Joseph S. Myers
joseph@codesourcery.com