r198858 - Disable LeakSanitizer in TableGen binaries, see PR18325

[Kostya Serebryany]

On Tue, Jan 14, 2014 at 1:47 AM, Richard Smith <richardsmith at google.com>wrote:

> On Mon, Jan 13, 2014 at 8:03 AM, Alp Toker <alp at nuanti.com> wrote:
>
>>
>> On 13/01/2014 15:02, Kostya Serebryany wrote:
>>
>>> Alp, others,
>>>
>>> Now that there is no urgent issue I'd like to return to the discussion
>>> about names with "__" prefix in sanitizer run-times.
>>>
>>
>> If you're OK to conditionalize the code so it's only defined in sanitizer
>> builds, that's fine and as Sean pointed out there's already plenty of
>> precedent -- it's what every software project does to define "__" prefixed
>> weak link functions like libc malloc hooks.
>>
>
> This discussion is so vastly far from the best use of any of our time.
> Let's just restore the name we had before and wrap the definition in
> __has_feature(leak_sanitizer). I think that will make everyone happy. OK?
>

There is no __has_feature(leak_sanitizer) and it can't be implemented with
reasonable semantics (discussed here before).
Unless someone objects, I'll change the code back to use
__lsan_is_turned_off and hide it under __has_feature(address_sanitizer),
then I will revert my change that added an alternative to
__lsan_is_turned_off (LeakSanitizerIsTurnedOffForTheCurrentProcess)

This means that  -Wreserved, -Wreserved-macros, -Wreserved-identifiers
(once implemented)
will not work on clang bootstrap in AddressSanitizer mode,
but let Alp handle this if it ever becomes a problem.

--kcc
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140114/8e50186e/attachment.html>