Always exclude ELF dynamic linker/loader

[mayeut]

s390x/ppc64le ELF dynamic linkers do not start with ld-linuxand therefore, were not excluded from analysis.

Relates to #212
Fixes pypa/manylinux#412

The tests did not reveal this issue and shall be updated. I don't know how to get this done since all symbols in the linker are private (or seem to be).

@wdirons, @st-pasha, any simple test we could include to reproduce this behavior in tests ?

[codecov]

Codecov Report

Merging #213 into master will decrease coverage by 2.09%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master     #213     +/-   ##
=========================================
- Coverage   89.25%   87.16%   -2.1%     
=========================================
  Files          19       19             
  Lines         996      974     -22     
  Branches      215      214      -1     
=========================================
- Hits          889      849     -40     
- Misses         65       86     +21     
+ Partials       42       39      -3

Impacted Files	Coverage Δ
auditwheel/policy/external_references.py	97.61% <100%> (ø)	⬆️
auditwheel/main_addtag.py	29.41% <0%> (-57.44%)	⬇️
auditwheel/main_repair.py	72.09% <0%> (-4.51%)	⬇️
auditwheel/wheeltools.py	89.51% <0%> (-2.42%)	⬇️
auditwheel/lddtree.py	92.14% <0%> (-0.72%)	⬇️
auditwheel/wheel_abi.py	98.31% <0%> (-0.09%)	⬇️
auditwheel/tmpdirs.py	100% <0%> (ø)	⬆️
auditwheel/repair.py	83.33% <0%> (+2.15%)	⬆️
auditwheel/main_show.py	77.08% <0%> (+4.86%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 3b1d9ef...22f551f. Read the comment docs.

[di]

I added two integration tests for this using numpy wheels for s390x and ppc64le thanks to @wdirons in #212 -- if these aren't what we were looking for, please feel free to amend/remove my commit.

Also, I think the s390x build will probably fail here due to #228, but we should probably merge this anyways.

[di]

Tests are now passing, what do you think @mayeut?

[mayeut]

@wdirons, @di,
thanks for the test case.
I'd really like not to have 25MB+ stored in the repo. I'll try to see if I can get this shrunk a bit.

[wdirons]

I'm trying to figure out how to modify the test_elfutils.py test for this instance.

I have noticed, the existing testcase for ld-linux is not testing what it should as this change still passes the test:

diff --git a/tests/unit/test_elfutils.py b/tests/unit/test_elfutils.py
index 23e31e7..045ad5b 100644
--- a/tests/unit/test_elfutils.py
+++ b/tests/unit/test_elfutils.py
@@ -99,7 +99,7 @@ class TestElfFindVersionedSymbols:
         # GIVEN
         elf = Mock()
         verneed = Mock()
-        verneed.configure_mock(name="ld-linux")
+        verneed.configure_mock(name="foo-bar")
         elf.get_section_by_name.return_value.iter_versions.return_value = (
             (verneed, []),
         )

[mayeut]

I finally managed to get some time to find a small test case.
Relying on TLS creates a dependency on the ELF dynamic linker/loader.

Failures can be seen here when the fix is not here: https://travis-ci.com/github/mayeut/auditwheel/jobs/320711188 & https://travis-ci.com/github/mayeut/auditwheel/jobs/320711187

[mayeut]

CI not passing but passing here https://travis-ci.com/github/mayeut/auditwheel/builds/160833993
so going ahead with merge