Fix dependencies on png.h, m4ri.h, m4rie.h

[Jeroen Demeyer]

The following dependencies were incorrect:

• sage.matrix.matrix_mod2_dense does not use png.h
• sage.modules.vector_mod2_dense does not use png.h
• sage.matrix.matrix_rational_dense does not use m4ri.h

The following dependencies are in Cython files, so Cython picks them up automatically:

• sage.matrix.matrix_mod2_dense dependency on m4ri.h
• sage.modules.vector_mod2_dense dependency on m4ri.h
• sage.matrix.matrix_gf2e_dense dependency on m4rie.h

The following include and dependency was removed:

• pb_wrap.h included m4ri/m4ri.h for no apparent reason, everything still works after removing that include (and therefore also the dependency)

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​ca7bfae

Fix dependencies on png.h, m4ri.h, m4rie.h

[Erik Bray]

If Cython can pick up such header dependencies automatically, why do we still also need the depends for sage.rings.polynomial.pbori?

[Jeroen Demeyer]

Replying to embray:

If Cython can pick up such header dependencies automatically, why do we still also need the depends for sage.rings.polynomial.pbori?

Cython picks up dependencies automatically if they appear in a cdef extern from "header.h"

But in this case, the dependency is through pb_wrap.h.

[Erik Bray]

I see. I suppose in principle one could put an empty cdef extern from "polybori/polybori.h" in the Cython file then. I would still like to see the use of SAGE_INC there go away, but that can be handled in #28349 or a different ticket.

[Jeroen Demeyer]

Replying to embray:

I suppose in principle one could put an empty cdef extern from "polybori/polybori.h" in the Cython file then.

Indeed, I was thinking the same thing.