disable test on x86-32 since compiletest is not flexible enough to express where this test can run

Author: RalfJung

tests/codegen/simd-intrinsic/simd-intrinsic-transmute-array.rs

@@ -2,10 +2,11 @@
 //@ compile-flags: -C no-prepopulate-passes
 // LLVM IR isn't very portable and the one tested here depends on the ABI
 // which is different between x86 (where we use SSE registers) and others
-//@ revisions: x86-32 x86-64 other
-// The two x86 revisions are identical. Ideally we would say "x86-32 or x86-64" in a single revision
-// but compiletest does not permit that.
-//@[x86-32] only-x86
+//@ revisions: x86-64 other
+// FIXME: there should be an x86-32 revision, but that doesn't work since
+// compiletest cannot distinguish i686 from i586 targets, and those produce
+// different code since their ABI differs and the tests below inadvertantly
+// are also testing ABI details.
 //@[x86-64] only-x86_64
 //@[other] ignore-x86
 //@[other] ignore-x86_64