[gpsd-dev] [PATCH 1/2] For numpy, use longdouble rather than float128

[Sanjeev Gupta]

On 32-bit machines, float128 may not be defined.

Additionally, there is no guarantee that float96
or float128, etc, will be defined or available on _any_
architecture.  The float128 and float96 exports are,
at best, 80-bit represenations with 48 and 16 bits of
padding.  longdouble is portable.
---
 contrib/ntpshmviz | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/contrib/ntpshmviz b/contrib/ntpshmviz
index c0b32d79..684798f2 100755
--- a/contrib/ntpshmviz
+++ b/contrib/ntpshmviz
@@ -93,8 +93,8 @@ class ntpOffset(object):
                     line = line.lstrip()
                     record = line.split(' ')
                     try:
-                        offset = (numpy.float128(record[3]) -
-                                  numpy.float128(record[4]))
+                        offset = (numpy.longdouble(record[3]) -
+                                  numpy.longdouble(record[4]))
                     except:
                         print ("Invalid data: ", sys.exc_info()[0],
                                ". Data was: ", line)
-- 
2.20.1