[cfe-commits] r172087 - in /cfe/trunk: test/TableGen/ test/TableGen/DiagnosticBase.inc test/TableGen/anonymous-groups.td test/TableGen/lit.local.cfg test/TableGen/tg-fixits.td utils/TableGen/ClangDiagnosticsEmitter.cpp

Thu Jan 10 17:44:05 PST 2013

On Jan 10, 2013, at 16:38 , Dmitri Gribenko <gribozavr at gmail.com> wrote:

> On Fri, Jan 11, 2013 at 12:25 AM, Dmitri Gribenko <gribozavr at gmail.com> wrote:
>> On Fri, Jan 11, 2013 at 12:20 AM, Jordan Rose <jordan_rose at apple.com> wrote:
>>> 
>>> On Jan 10, 2013, at 14:19 , Dmitri Gribenko <gribozavr at gmail.com> wrote:
>>> 
>>>> On Thu, Jan 10, 2013 at 11:54 PM, Jordan Rose <jordan_rose at apple.com> wrote:
>>>>> 
>>>>> On Jan 10, 2013, at 13:51 , Dmitri Gribenko <gribozavr at gmail.com> wrote:
>>>>> 
>>>>> On Thu, Jan 10, 2013 at 11:47 PM, Jordan Rose <jordan_rose at apple.com> wrote:
>>>>> 
>>>>> 
>>>>> On Jan 10, 2013, at 13:41 , Dmitri Gribenko <gribozavr at gmail.com> wrote:
>>>>> 
>>>>> On Thu, Jan 10, 2013 at 8:50 PM, Jordan Rose <jordan_rose at apple.com> wrote:
>>>>> 
>>>>> --- cfe/trunk/test/TableGen/tg-fixits.td (added)
>>>>> +++ cfe/trunk/test/TableGen/tg-fixits.td Thu Jan 10 12:50:46 2013
>>>>> @@ -0,0 +1,41 @@
>>>>> +// RUN: clang-tblgen -gen-clang-diag-groups -I%S %s -o /dev/null 2>&1 |
>>>>> FileCheck --strict-whitespace %s
>>>>> +include "DiagnosticBase.inc"
>>>>> +
>>>>> +def NamedGroup : DiagGroup<"name">;
>>>>> +
>>>>> +def InNamedGroup : Warning<"">, InGroup<DiagGroup<"name">>;
>>>>> +//      CHECK: tg-fixits.td:[[@LINE-1]]:41: error: group 'name' is referred
>>>>> to anonymously
>>>>> +// CHECK-NEXT: {{^def InNamedGroup : Warning<"">,
>>>>> InGroup<DiagGroup<"name">>;}}
>>>>> +// CHECK-NEXT: {{^
>>>>> ~~~~~~~~\^~~~~~~~~~~~~~~~~~}}
>>>>> +// CHECK-NEXT: {{^                                InGroup<NamedGroup>}}
>>>>> +
>>>>> +def Wrapped : Warning<"">, InGroup<DiagGroup<
>>>>> +  "name">>;
>>>>> +//      CHECK: tg-fixits.td:[[@LINE-2]]:36: error: group 'name' is referred
>>>>> to anonymously
>>>>> +// CHECK-NEXT: {{^def Wrapped : Warning<"">, InGroup<DiagGroup<}}
>>>>> +// CHECK-NEXT: {{^                           ~~~~~~~~\^~~~~~~~~~}}
>>>>> +// CHECK-NEXT: {{^                           InGroup<NamedGroup>}}
>>>>> +
>>>>> +def AlsoWrapped : Warning<"">, InGroup<
>>>>> +  DiagGroup<"name">>;
>>>>> +//      CHECK: tg-fixits.td:[[@LINE-1]]:3: error: group 'name' is referred
>>>>> to anonymously
>>>>> +// CHECK-NEXT: {{^  DiagGroup<"name">>;}}
>>>>> +// CHECK-NEXT: {{^~~\^~~~~~~~~~~~~~~~~~}}
>>>>> +// CHECK-NEXT: {{^InGroup<NamedGroup>}}
>>>>> +
>>>>> +// The following lines contain hard tabs (\t); do not change this!
>>>>> +def HardTabs : Warning<"">,
>>>>> +       InGroup<        DiagGroup<"name">       >;
>>>>> +//      CHECK: tg-fixits.td:[[@LINE-1]]:11: error: group 'name' is referred
>>>>> to anonymously
>>>>> +// CHECK-NEXT: {{^        InGroup<        DiagGroup<"name">       >;}}
>>>>> +// CHECK-NEXT: {{^        ~~~~~~~~~~~~~~~~\^~~~~~~~~~~~~~~~~~~~~~~~~}}
>>>>> +// CHECK-NEXT: {{^        InGroup<NamedGrop>}}
>>>>> +
>>>>> +// The following line has Unicode characters in it; do not change them!
>>>>> +// FIXME: For now, we just give up on printing carets/ranges/fixits for
>>>>> +// lines with Unicode in them, because SMDiagnostic don't keep a
>>>>> byte<->column
>>>>> +// map around to line things up like Clang does.
>>>>> +def Unicode : Warning<"ãƒ¦ãƒ‹ã‚³ãƒ¼ãƒ‰">, InGroup<DiagGroup<"name">>;
>>>>> +//      CHECK: tg-fixits.td:[[@LINE-1]]:51: error: group 'name' is referred
>>>>> to anonymously
>>>>> +// CHECK-NEXT: def Unicode : Warning<"{{[^"]+}}">,
>>>>> InGroup<DiagGroup<"name">>;
>>>>> +// CHECK-NEXT: note:
>>>>> 
>>>>> 
>>>>> Hello Jordan,
>>>>> 
>>>>> The Unicode test fails:
>>>>> http://lab.llvm.org:8011/builders/clang-x86_64-debian-fast/builds/119
>>>>> 
>>>>> Dmitri
>>>>> 
>>>>> 
>>>>> Hm. That doesn't make sense to me; the Unicode is clearly in the output,
>>>>> Japanese characters require three bytes of UTF-8, and presumably take 1 or 2
>>>>> columns each. Is there a way to get the input that FileCheck is seeing from
>>>>> this builder?
>>>>> 
>>>>> 
>>>>> The Japanese characters are fine (but since they are double-width, our
>>>>> underlining in the terminal is funny).  The issue is with missing
>>>>> CHECK lines.
>>>>> 
>>>>> 
>>>>> Right, but the code in SourceMgr.cpp is supposed to say "oh, we can't print
>>>>> a caret here because we'll get it wrong", and leave out the caret and fixit
>>>>> lines (see LLVM r172086). Why would that fail?
>>>> 
>>>> Breakpoint 1, llvm::SMDiagnostic::print (this=0x7fffffffd408,
>>>> ProgName=0x0, S=..., ShowColors=true) at
>>>> /home/gribozavr/clang/llvm/lib/Support/SourceMgr.cpp:400
>>>> 400     SmallString<128> PrintableLine(LineContents);
>>>> (gdb) p LineContents.data()
>>>> $3 = 0x78cb38 "def Unicode : Warning<\"ユニコード\">, InGroup<DiagGroup<\"name\">>;"
>>>> (gdb) n
>>>> 401     std::replace(PrintableLine.begin(), PrintableLine.end(), '\t', ' ');
>>>> (gdb) n
>>>> 402     size_t NumColumns = (size_t)llvm::sys::locale::columnWidth(PrintableLine);
>>>> (gdb) n
>>>> 403     if (NumColumns != PrintableLine.size()) {
>>>> (gdb) p NumColumns
>>>> $4 = 69
>>>> (gdb) p PrintableLine.size()
>>>> $5 = 69
>>>> 
>>>> Seems like columnWidth() is not working properly in this case.  I'm
>>>> running with LANG=ru_UA.utf8, same results with LANG=en_US.utf8.
>>> 
>>> Great. Okay, can you file a PR that columnWidth doesn't seem to work correctly on Linux systems? Meanwhile, I'll just blanket this check to "any non-ASCII characters" for now.
>> 
>> Of course it is not working! :)  LocaleGeneric is used for some reason:
>> 
>> (gdb) b llvm::sys::locale::columnWidth
>> Breakpoint 1 at 0x5186b4: file
>> /home/gribozavr/clang/llvm/lib/Support/LocaleGeneric.inc, line 8.
>> 
>> I will file a PR.
> 
> Fixed up the test in r172152.
> 
> http://llvm.org/bugs/show_bug.cgi?id=14910

Sorry, this is wrong. The test is correct on OS X (and presumably Windows as well). I'm working on another patch that just bans multibyte characters rather than checking column widths.