eval: Fix int-to-float conversions

Also, add bounds checks for float-to-int conversions. If the integer
part can't be represented in the result type, C behavior is undefined.
Although this means the result is arbitrary, we need to avoid
undefined behavior in cproc itself when given such a program as
input.

4 files changed, 4 insertions, 0 deletions

diff --git a/test/initializer-cast-float-int.c b/test/initializer-cast-float-int.c
new file mode 100644
index 0000000..42296e9
--- /dev/null
+++ b/test/initializer-cast-float-int.c
@@ -0,0 +1 @@
+int x = -1.0;
diff --git a/test/initializer-cast-float-int.qbe b/test/initializer-cast-float-int.qbe
new file mode 100644
index 0000000..675fc96
--- /dev/null
+++ b/test/initializer-cast-float-int.qbe
@@ -0,0 +1 @@
+export data $x = align 4 { w 18446744073709551615, }
diff --git a/test/initializer-cast-int-float.c b/test/initializer-cast-int-float.c
new file mode 100644
index 0000000..e041d90
--- /dev/null
+++ b/test/initializer-cast-int-float.c
@@ -0,0 +1 @@
+double x = -1;
diff --git a/test/initializer-cast-int-float.qbe b/test/initializer-cast-int-float.qbe
new file mode 100644
index 0000000..80b905e
--- /dev/null
+++ b/test/initializer-cast-int-float.qbe
@@ -0,0 +1 @@
+export data $x = align 8 { d d_-1, }