[lld] r257661 - Attempt to make FreeBSD buildbot green.

[Davide Italiano via llvm-commits]

On Wed, Jan 13, 2016 at 1:13 PM, Ed Maste <emaste at freebsd.org> wrote:
> On 13 January 2016 at 20:14, Davide Italiano <davide at freebsd.org> wrote:
>> On Wed, Jan 13, 2016 at 11:40 AM, Rui Ueyama via llvm-commits
>> <llvm-commits at lists.llvm.org> wrote:
>>> Author: ruiu
>>> Date: Wed Jan 13 13:40:13 2016
>>> New Revision: 257661
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=257661&view=rev
>>> Log:
>>> Attempt to make FreeBSD buildbot green.
>>>
>>> It seems that __cxa_demangle function on the buildbot tried to demangle
>>> a variable "tlsvar" as a C++ symbol. Do not call that function unless
>>> it does not start with "_Z" which is the prefix of mangled names.
>>>
>>> Modified:
>>>     lld/trunk/ELF/Symbols.cpp
>>>
>>> Modified: lld/trunk/ELF/Symbols.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/lld/trunk/ELF/Symbols.cpp?rev=257661&r1=257660&r2=257661&view=diff
>>> ==============================================================================
>>> --- lld/trunk/ELF/Symbols.cpp (original)
>>> +++ lld/trunk/ELF/Symbols.cpp Wed Jan 13 13:40:13 2016
>>> @@ -145,6 +145,11 @@ std::string elf2::demangle(StringRef Nam
>>>  #else
>>>    if (!Config->Demangle)
>>>      return Name;
>>> +
>>> +  // Return if it does not look like a C++ symbol.
>>> +  if (!Name.startswith("_Z"))
>>> +    return Name;
>>> +
>>>    char *Buf =
>>>        abi::__cxa_demangle(Name.str().c_str(), nullptr, nullptr, nullptr);
>>>    if (!Buf)
>>>
>>
>> This seems to be a bug in libcxxabi.
>
> If __cxa_demangle is being used to demangle external names the _Z test
> is correct, and required. __cxa_demangle can also be used to demangle
> types, and will for example correctly return "float" for input "f".
>
> On FreeBSD __cxa_demangle is provided by libcxxrt, and this issue has
> come up a few times in the past. libcxxrt's __cxa_demangle ought to
> reject names with additional characters following a mangled string,
> but it becomes much less of an issue once callers are fixed to add the
> _Z test.
>
> Some previous instances:
> https://lists.freebsd.org/pipermail/freebsd-toolchain/2014-June/001153.html
> http://reviews.llvm.org/D2552

Thank you. I read the thread and I agree (therefore disagreeing with
David C.) that this is an hack. Probably not a terrible one, but
ideally lld (and all the other tools) shouldn't be concerned with this
particular bit of the ABI.

-- 
Davide

"There are no solved problems; there are only problems that are more
or less solved" -- Henri Poincare