The combination of longer timeout and larger number of retries should
make the linkcheck sufficiently specific again.

Retrying > once should increase the specificity of the check a lot.

Fixup/qsymm tests

See merge request kwant/kwant!330

sympy symbols with 'real' set differently do not hash to the same
value (and possibly do not compare equal).

__eq__ for Models just uses allclose anyway.

This is more idiomatic.

Also comment other usages of this parameter.

Qsymm is now installed in the testing image directly

Not providing these is now deprecated.
Close #325.

fix kpm.conductivity in the case where 'norbs' is not set

See merge request kwant/kwant!321

This does not require assuming that site families have 'norbs' set.

Closes #313.

fix a bug in kwant.qsymm

See merge request kwant/kwant!325

The test suite uses features from the 1.2 release of qsymm.

This is not an API of qsymm.

Sphinx 2 got rid of 'sphinx.main' so we can no longer use it.

In pytest v5 the behaviour of str(exc_info) changed. We now
use str(exc_info.value) as the canonical way of obtaining
the exception message.

Previously when 'attach_lead' would expand a lead unit cell to deal
with greater than nearest-neighbor hoppings it would unset the
conservation law and discrete symmetries of the lead.

Closes #303.

Closes #279.

Closes #214.

Closes #295.

The function which finds the symmetries of kwant builders does not find all real space symmetries if the unit cell contains more than one sites, because the conversion from builder to qsymm model does not use a basis which preserves information on the real space position of sites. The tests do not catch this, because they only have one site per unit cell. This fixes the basis choice, and adds a test.

Since this does not seem to justify a new bugfix release of Kwant 1.3,
I apply this commit to the 1.4 stable branch.

When System.parameters was introduced, we thought it would be a good idea to
also use this new attribute.  The plan was to check in hamiltonian_submatrix()
whether any unneeded parameters are provided by the user.  This turned out to
be impossible, so the only check that remained was whether all the necessary
parameters are indeed provided.  However, this check is not necessary, because
the 'hamiltonian' method is in any way obliged to signal an error when it does
not obtain the necessary parameters.