From 5d02c918d6f8ae9f4d0c5698d01c8d05b3d63f79 Mon Sep 17 00:00:00 2001
From: Pavel Raiskup <praiskup@redhat.com>
Date: Jan 25 2017 16:32:48 +0000
Subject: multilib-fix: document why not to use __WORDSIZE


Fedora's PackagingDrafts/MultilibTricks suggests the following
c-header wrapper example:

  #include <bits/wordsize.h>
  #if __WORDSIZE == 32
  #include "config-32.h"
  #elif __WORDSIZE == 64
  #include "config-64.h"
  #else
  #error "Unknown word size"
  #endif

Installing and using this wrapper is not really no-op, however; it
has been proven to cause include-ordering issues.  So, even though
the shorter stub works in most cases, document why we still prefer
the longer #ifdef wrapper.  Thanks to Vít Ondruch for pointing
this out.

Related: rhbz#1412274

---

diff --git a/multilib-fix b/multilib-fix
index eafb0b7..254d84a 100755
--- a/multilib-fix
+++ b/multilib-fix
@@ -48,6 +48,10 @@ cat <<EOF
  * To avoid breaking arches not explicitly supported by Red Hat, we
  * use this indirection file *only* on known multilib arches.
  *
+ * We pay attention to include _only_ the original multilib-unclean
+ * header file.  Including any other system-header file could cause
+ * unpredictable include-ordering issues (rhbz#1412274, comment #16).
+ *
  * Note: this may well fail if user tries to use gcc's -I- option.
  * But that option is deprecated anyway.
  */