From 933e701dac1936c6f15c765b5687fbc623464ec7 Mon Sep 17 00:00:00 2001
From: Mike Sperber <sperber@deinprogramm.de>
Date: Thu, 22 Mar 2012 18:00:31 +0100
Subject: Unbreak floating point on middle-endian machines.

On some ARMs (and maybe other platforms), doubles are stored with the
the two 32-bit words reversed with respect to more common
architectures.

The symptom is this:

> io_lib:write(1.0).
"0.000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000005299808824"

Detect that and account for it when decoding floats.
---
 erts/configure.in | 1 +
 1 file changed, 1 insertion(+)

(limited to 'erts/configure.in')

diff --git a/erts/configure.in b/erts/configure.in
index b801994e14..6583db8116 100644
--- a/erts/configure.in
+++ b/erts/configure.in
@@ -1681,6 +1681,7 @@ case X$erl_xcomp_bigendian in
 esac
 
 AC_C_BIGENDIAN
+AC_C_DOUBLE_MIDDLE_ENDIAN
 
 dnl fdatasync syscall (Unix only)
 AC_CHECK_FUNCS([fdatasync])
-- 
cgit v1.2.3