hide the ASCII_ARMOR symbol

There appears to be some interaction between once_cell and the 32-bit
toolchain for windows that is causing the ASCII_ARMOR symbol to be
mis-linked. To work around the issue reported in #29, hide the
ASCII_ARMOR symbol inside of a function so that it is not even
crate-wide available but is only available inside the function.

CHANGELOG.md

@@ -1,3 +1,7 @@
+# 1.0.1
+
+ - hide the ASCII_ARMOR symbol to work around a linking issue with 32-bit windows builds
+
 # 1.0
 
  - `pem::parse_many` now returns a `Result<Vec<Pem>>` instead of a `Vec<Pem>` that silently discarded invalid sections.

src/lib.rs

@@ -116,9 +116,14 @@ const REGEX_STR: &str =
 /// The line length for PEM encoding
 const LINE_WRAP: usize = 64;
 
-static ASCII_ARMOR: Lazy<Regex> = Lazy::new(|| {
-    Regex::new(REGEX_STR).unwrap()
-});
+
+fn ascii_armor() -> &'static Regex {
+    static ASCII_ARMOR: Lazy<Regex> = Lazy::new(|| {
+        Regex::new(REGEX_STR).unwrap()
+    });
+
+    &ASCII_ARMOR
+}
 
 /// Enum describing line endings
 #[derive(Debug, Clone, Copy)]
@@ -242,7 +247,7 @@ impl Pem {
 ///  assert_eq!(pem.tag, "RSA PRIVATE KEY");
 /// ```
 pub fn parse<B: AsRef<[u8]>>(input: B) -> Result<Pem> {
-    ASCII_ARMOR
+    ascii_armor()
         .captures(&input.as_ref())
         .ok_or_else(|| PemError::MalformedFraming)
         .and_then(Pem::new_from_captures)
@@ -319,7 +324,7 @@ pub fn parse<B: AsRef<[u8]>>(input: B) -> Result<Pem> {
 /// ```
 pub fn parse_many<B: AsRef<[u8]>>(input: B) -> Result<Vec<Pem>> {
     // Each time our regex matches a PEM section, we need to decode it.
-    ASCII_ARMOR
+    ascii_armor()
         .captures_iter(&input.as_ref())
         .map(|caps| Pem::new_from_captures(caps))
         .collect()