Modulo a few comments, I think this looks fine, but I don't think I agree with 
using this to mark tests that fail under a given sanitizer. Specifically, I 
don't think "unsupported" is semantically correct in this case. If the test is 
failing under a given sanitizer, there's either something wrong with the test, 
something wrong with one of our libraries, or something wrong with the 
sanitizer itself, right? In that case, I think the right thing to do would be 
to XFAIL them under those conditions.

On a somewhat related note, I noticed that the [LIT 
docs](http://llvm.org/docs/TestingGuide.html#other-features) claim we're 
supposed to be able to label XFAILs with a PR number, but a few quick attempts 
with this didn't show any useful feedback in the test results. I assume this is 
because we parse the XFAIL lines ourselves. We should fix this so we can track 
these better.

================
Comment at: test/lit.cfg:91
@@ -87,1 +90,3 @@
+                                     if f not in test.config.available_features
+                                    and f not in test.config.target_triple]
         if missing_required_features:
----------------
Whitespace alignment.

Also, what parts of the test suite would need REQUIRES: <target-triple>? I 
don't think I've found any tests that are platform specific. If not, YAGNI.

http://reviews.llvm.org/D4950



_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to