[PATCH] D17643: Fix PR26410: failing assertion when the X87 stackifier runs on functions containing IRETs

[Quentin Colombet via llvm-commits]

qcolombet added a comment.

Hi David,

The change LGTM, but the tests need a few nitpicks.

Cheers,
-Quentin


================
Comment at: test/CodeGen/X86/x86-32-intrcc.ll:83
@@ +82,3 @@
+; Test that the presence of x87 does not crash the FP stackifier
+define x86_intrcc void @test_isr_x87(%struct.interrupt_frame* %frame) {
+entry:
----------------
Could you check we generate sensible code here?

================
Comment at: test/CodeGen/X86/x86-32-intrcc.ll:86
@@ +85,3 @@
+  %0 = load x86_fp80, x86_fp80* @f80, align 4
+  %add = fadd x86_fp80 %0, 0xK3FFF8000000000000000
+  store x86_fp80 %add, x86_fp80* @f80, align 4
----------------
Please give a name to %[0-9]+ variables.
opt -instnamer can do this for you.
(It is a pain to update test cases with numbered variables.)

================
Comment at: test/CodeGen/X86/x86-64-intrcc.ll:90
@@ +89,3 @@
+
+; Test that the presence of x87 does not crash the FP stackifier
+define x86_intrcc void @test_isr_x87(%struct.interrupt_frame* %frame) {
----------------
Ditto: no %[0-9]+ and sensible testing.


http://reviews.llvm.org/D17643