In the mat2cell docstring header, the variable "d?" has been renamed to "dim?" in changeset https://hg.savannah.gnu.org/hgweb/octave/rev/b326eeae92d0
but in the text, there is still one "di" left.
Here is a patch for this:

--- a/libinterp/corefcn/cellfun.cc
+++ b/libinterp/corefcn/cellfun.cc
@@ -2112,7 +2112,7 @@
 integers which specifies how to divide that dimension's elements amongst the
 new elements in the output @var{C}.  The number of elements in the @var{i}-th
 dimension is @code{size (@var{A}, @var{i})}.  Because all elements in @var{A}
-must be partitioned, there is a requirement that @code{sum (@var{di}) == size
+must be partitioned, there is a requirement that @code{sum (@var{dimi}) == size
 (@var{A}, i)}.  The size of the output cell @var{C} is numel (@var{dim1}) x
 numel (@var{dim2}) x @dots{} x numel (@var{dimn}).

And here is another small patch for polyeig.m

--- a/scripts/polynomial/polyeig.m
+++ b/scripts/polynomial/polyeig.m
@@ -29,8 +29,8 @@
 ##
 ## @code{polyeig} solves the eigenvalue problem
 ##
-## @code{(@var{C0} + @var{C1} @var{z} + @dots{} + @var{Cl} z^@var{l}) @var{v} =
-## 0}.
+## @code{(@var{C0} + @var{C1} @var{z} + @dots{} + @var{Cl} @var{z}^@var{l})
+## @var{v} = 0}.
 ##
 ## Note that the eigenvalues @var{z} are the zeros of the matrix polynomial.
 ## @var{z} is a row vector with @code{@var{n}*@var{l}} elements.  @var{v} is a