r215046 - Use -Rblah, not -Wblah, to control remark diagnostics. This was always the

[Justin Bogner]

David Blaikie <dblaikie at gmail.com> writes:
> On Thu, Aug 7, 2014 at 11:57 AM, Justin Bogner <mail at justinbogner.com> wrote:
>> Richard Smith <richard-llvm at metafoo.co.uk> writes:
>>> Author: rsmith
>>> Date: Wed Aug  6 19:24:21 2014
>>> New Revision: 215046
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=215046&view=rev>> Log:
>>> Use -Rblah, not -Wblah, to control remark diagnostics. This was always the
>>> intent when we added remark support, but was never implemented in the general
>>> case, because the first -R flags didn't need it. (-Rpass= had
>>> special handling
>>> to accomodate its argument.)
>>
>> <snip>
>>
>>> Modified: cfe/trunk/test/Modules/Rmodule-build.m
>>> URL:
>>> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Modules/Rmodule-build.m?rev=215046&r1=215045&r2=215046&view=diff==============================================================================>> --- cfe/trunk/test/Modules/Rmodule-build.m (original)
>>> +++ cfe/trunk/test/Modules/Rmodule-build.m Wed Aug  6 19:24:21 2014
>>> @@ -7,7 +7,7 @@
>>>  // RUN: echo 'module B { header "B.h" }' >> %t/module.modulemap
>>>
>>>  // RUN: %clang_cc1 -fmodules -fmodules-cache-path=%t -fsyntax-only
>>> %s -verify \
>>> -// RUN:            -I %t -Wmodule-build
>>> +// RUN:            -I %t -Rmodule-build
>>>
>>>  @import A; // expected-remark{{building module 'A' as}}
>>>  @import B; // expected-remark{{building module 'B' as}}
>>> @@ -16,7 +16,19 @@
>>>
>>>  // RUN: echo ' ' >> %t/B.h
>>>  // RUN: %clang_cc1 -fmodules -fmodules-cache-path=%t -fsyntax-only
>>> %s -I %t \
>>> -// RUN:            -Wmodule-build 2>&1 | FileCheck %s
>>> +// RUN:            -Rmodule-build 2>&1 | FileCheck %s
>>> +
>>> +// RUN: echo ' ' >> %t/B.h
>>> +// RUN: %clang_cc1 -fmodules -fmodules-cache-path=%t -fsyntax-only
>>> %s -I %t \
>>> +// RUN:            -Reverything 2>&1 | FileCheck %s
>>> +
>>> +// RUN: echo ' ' >> %t/B.h
>>> +// RUN: %clang_cc1 -fmodules -fmodules-cache-path=%t -fsyntax-only
>>> %s -I %t \
>>> +// RUN:            2>&1 | count 0
>>> +
>>> +// RUN: echo ' ' >> %t/B.h
>>> +// RUN: %clang_cc1 -fmodules -fmodules-cache-path=%t -fsyntax-only
>>> %s -I %t \
>>> +// RUN:            -Rmodule-build -Rno-everything 2>&1 | count 0
>>>
>>>  // CHECK-NOT: building module 'A'
>>>  // CHECK: building module 'B'
>>
>> stderr piped into "count 0" in tests breaks when guard malloc is
>> involved. Please don't do it.
>
> I'm curious - why does guard malloc hurt this? I imagine anything that
> just writes extra stuff to stdout/stderr is probably not something we
> want in the test environment (it sounds like that's what you're
> implying guard malloc does).

It writes out something like so on stderr when it starts. I guess it's
kind of weird if it's doing that when nothing's wrong, but c'est la vie.

    GuardMalloc[clang-41241]: Allocations will be placed on 16 byte boundaries.
    GuardMalloc[clang-41241]:  - Some buffer overruns may not be noticed.
    GuardMalloc[clang-41241]:  - Applications using vector instructions (e.g., SSE) should work.
    GuardMalloc[clang-41241]: version 25

>
>>
>> Historically there hasn't been a very good alternative, so I've added a
>> -allow-empty flag to FileCheck in r215127 and used it in r215129 to fix
>> this.
>> _______________________________________________
>> cfe-commits mailing list
>> cfe-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits