[cfe-commits] r161501 - in /cfe/trunk: include/clang/AST/Expr.h include/clang/Basic/DiagnosticGroups.td include/clang/Basic/DiagnosticSemaKinds.td lib/AST/Expr.cpp lib/Sema/SemaExpr.cpp lib/Sema/SemaOverload.cpp test/SemaCXX/constant-expression-c

[David Dean]

On 8 Aug 2012, at 5:13 PM, David Blaikie <dblaikie at gmail.com> wrote:

> On Wed, Aug 8, 2012 at 4:37 PM, David Dean <david_dean at apple.com> wrote:
>> Some of the dejagnu tests are failing after this change, can you verify that the output is correct and update the test cases?
>> 
>> g++.dg/conversion/nullptr1.C (test for excess errors)
>> g++.dg/init/null1.C (test for excess errors)
>> g++.dg/init/self1.C (test for excess errors)
> 
> Is there a sensible way to run this on a Linux machine or do I need to
> perform various hackery on the scripts and/or have an Apple box handy?

I would expect that dejagnu behaves reasonably well on linux, but I haven't tried it.

> Would it be easier for you to send me some logs & I can say yay or
> nay? I can't quite tell (not being familiar with dejagnu) what the
> current state of these tests are/where the expected output is compared
> against the actual output.

I'll send the logs and some brief commentary on them privately.


> But in all 3 cases there's an "interesting" null pointer that my
> diagnostic does catch & it's not unreasonable for it to do so, in my
> opinion:

I wasn't worried about false positives, but I was worried about the case you mention here:

> The 3rd one actually has a "dg-error" comment on it, I'm not sure if
> that's meant to be verifying that the compiler produces a diagnostic
> on this line of code, but I don't think clang did diagnose that until
> my change was made.

I'll double check previous logs to see if was triggering a diagnostic or not. If it was producing a diagnostic before, should we now have both, or does the one satisfy both?


-David