[pypy-commit] cffi add-float-dotdotdot: Document

[arigo]

Author: Armin Rigo <arigo at tunes.org>
Branch: add-float-dotdotdot
Changeset: r2261:f00569dace33
Date: 2015-08-28 11:01 +0200
http://bitbucket.org/cffi/cffi/changeset/f00569dace33/

Log:	Document

diff --git a/doc/source/cdef.rst b/doc/source/cdef.rst
--- a/doc/source/cdef.rst
+++ b/doc/source/cdef.rst
@@ -394,6 +394,16 @@
    ``(u)int(8,16,32,64)_t`` in Python, but in the generated C code,
    only ``foo_t`` is used.
 
+* *New in version 1.3:* floating-point types: "``typedef
+  float... foo_t;``" (or equivalently "``typedef double... foo_t;``")
+  declares ``foo_t`` as a-float-or-a-double; the compiler will figure
+  out which it is.  Note that if the actual C type is even larger
+  (``long double`` on some platforms), then compilation will fail.
+  The problem is that the Python "float" type cannot be used to store
+  the extra precision.  (Use the non-dot-dot-dot syntax ``typedef long
+  double foo_t;`` as usual, which returns values that are not Python
+  floats at all but cdata "long double" objects.)
+
 *  unknown types: the syntax "``typedef ... foo_t;``" declares the type
    ``foo_t`` as opaque.  Useful mainly for when the API takes and returns
    ``foo_t *`` without you needing to look inside the ``foo_t``.  Also
diff --git a/doc/source/whatsnew.rst b/doc/source/whatsnew.rst
--- a/doc/source/whatsnew.rst
+++ b/doc/source/whatsnew.rst
@@ -3,8 +3,12 @@
 ======================
 
 
-v1.2.2
-======
+v1.3
+====
+
+* Pull request #64: out-of-line API mode: we can now declare
+  floating-point types with ``typedef float... foo_t;``.  This only
+  works if ``foo_t`` is a float or a double, not ``long double``.
 
 * Issue #217: fix possible unaligned pointer manipulation, which crash
   on some architectures (64-bit, non-x86).