Issue #13545 has been reported by mame (Yusuke Endoh).

----------------------------------------
Bug #13545: Ruby built by clang 4.0.0 parses Float literal wrongly
https://bugs.ruby-lang.org/issues/13545

* Author: mame (Yusuke Endoh)
* Status: Open
* Priority: Normal
* Assignee: 
* Target version: 
* ruby -v: 
* Backport: 2.2: UNKNOWN, 2.3: UNKNOWN, 2.4: UNKNOWN
----------------------------------------
~~~
$ ./configure --prefix=/home/mame/work/local CC=clang
$ make -j 4 miniruby
$ ./miniruby -e 'p 1.1851510441583988e+30'
1.1851510441583981e+30
~~~

See the last digit.  While the original literal ends with "8e+30", the result ends with "1e+30".  The version of clang is 4.0.0-1ubuntu1.


There seems to be a bug in `ruby_strtod`.  It manipulates a double variable as an integer variable by using cast.  I think that this is undefined behavior, and clang seems to optimize some operations out.

We can fix this issue by adding `volatile` to its variable declaration, but I'm unsure if this is a good solution.

~~~
diff --git a/util.c b/util.c
index 4ae7c4fc96..ecd55bb0c5 100644
--- a/util.c
+++ b/util.c
@@ -1832,7 +1832,7 @@ d2b(double d_, int *e, int *bits)
 static double
 ratio(Bigint *a, Bigint *b)
 {
-    double_u da, db;
+    volatile double_u da, db;
     int k, ka, kb;
 
     dval(da) = b2d(a, &ka);
~~~



-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:ruby-core-request / ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>