fix internal dependencies of libraries

[louhibi]

Found a bug when two folders of dependencies have the same soname:
`pp external_refs_by_fn

{'libxml2mod.so': {'linux_x86_64': {'libs': {'libc.so.6': '/lib64/libc-2.5.so',
'libdl.so.2': '/lib64/libdl-2.5.so',
'libexslt.so.0': '/usr/local/lib/libexslt.so.0.8.7',
'libm.so.6': '/lib64/libm-2.5.so',
'libpthread.so.0': '/lib64/libpthread-2.5.so',
'libxml2.so.2': '/usr/local/lib/libxml2.so.2.6.21',
'libxslt.so.1': '/usr/local/lib/libxslt.so.1.1.9',
'libz.so.1': '/lib64/libz.so.1.2.3'},
'priority': 0},
'manylinux1_x86_64': {'libs': {'libexslt.so.0': '/usr/local/lib/libexslt.so.0.8.7',
'libxml2.so.2': '/usr/local/lib/libxml2.so.2.6.21',
'libxslt.so.1': '/usr/local/lib/libxslt.so.1.1.9',
'libz.so.1': '/lib64/libz.so.1.2.3'},
'priority': 100}},
'libxsltmod.so': {'linux_x86_64': {'libs': {'libc.so.6': '/lib64/libc-2.5.so',
'libdl.so.2': '/lib64/libdl-2.5.so',
'libexslt.so.0': '/usr/local/lib/libexslt.so.0.8.7',
'libm.so.6': '/lib64/libm-2.5.so',
'libpthread.so.0': '/lib64/libpthread-2.5.so',
'libxml2.so.2': '/usr/local/lib/libxml2.so.2.6.21',
'libxslt.so.1': '/usr/local/lib/libxslt.so.1.1.9',
'libz.so.1': '/lib64/libz.so.1.2.3'},
'priority': 0},
'manylinux1_x86_64': {'libs': {'libexslt.so.0': '/usr/local/lib/libexslt.so.0.8.7',
'libxml2.so.2': '/usr/local/lib/libxml2.so.2.6.21',
'libxslt.so.1': '/usr/local/lib/libxslt.so.1.1.9',
'libz.so.1': '/lib64/libz.so.1.2.3'},
'priority': 100}}}
`

and we endup with a soname_map that looks like this:

(Pdb) pp soname_map
{'libexslt.so.0': ('libexslt-bdf34a56.so.0.8.7',
'.libslibxsltmod/libexslt-bdf34a56.so.0.8.7'),
'libxml2.so.2': ('libxml2-e6a3f6dc.so.2.6.21',
'.libslibxsltmod/libxml2-e6a3f6dc.so.2.6.21'),
'libxslt.so.1': ('libxslt-88a6ba3e.so.1.1.9',
'.libslibxsltmod/libxslt-88a6ba3e.so.1.1.9'),
'libz.so.1': ('libz-a147dcb0.so.1.2.3',
'.libslibxsltmod/libz-a147dcb0.so.1.2.3')}

All the SOs in the .libslibxml2mod don't get updated.

[ehashman]

Can you please add a test to show the bug being fixed?

[louhibi]

@lpsinger will your PR #90 fix this bug ?

[louhibi]

@ehashman I will try to add a test if #90 will not fix this one too.

[lpsinger]

@lpsinger will your PR #90 fix this bug ?

Yes, it should.

[ehashman]

I will close this in favour of #90 (which I think is a more direct solution) and adding a test.