[cfe-dev] switch over constant results in strange constructor/destructor ordering

[Richard Smith via cfe-dev]

Thanks for letting us know, fixed in r281797.

On Fri, Sep 16, 2016 at 2:13 PM, Richard Smith <richard at metafoo.co.uk>
wrote:

> Wow, yes, that's a (pretty scary) wrong-code bug. Working on it...
>
> On Fri, Sep 16, 2016 at 6:47 AM, Andrew Johnson via cfe-dev <
> cfe-dev at lists.llvm.org> wrote:
>
>> Hello,
>>
>> I'm seeing an unexpected construction/destruction order in the program
>> below with clang 3.9.0. Is this considered a bug? Thank you.
>>
>> clang asm - https://godbolt.org/g/uGlpFa
>> gcc asm - https://godbolt.org/g/WDgcZP
>>
>> printf Output from clang (Unexpected):
>> obj entered 0
>> obj entered 1
>> obj entered 2
>> obj exited 2
>> obj exited 1
>> obj exited 0
>>
>> printf Output from gcc (Expected):
>> obj entered 0
>> obj exited 0
>> obj entered 1
>> obj exited 1
>> obj entered 2
>> obj exited 2
>>
>>
>> Program:
>>
>> #include <stdint.h>
>> #include <stdio.h>
>>
>> class SomeClass
>> {
>> uint32_t k;
>> public:
>> SomeClass(uint32_t x)
>>         { k = x; printf("obj entered %d\n", k); }
>> ~SomeClass()
>>         { printf("obj exited %d\n", k); }
>>
>> };
>>
>> int main()
>> {
>> uint32_t ii = 0;
>>
>> switch(1)
>> {
>> default:
>> {
>>
>> {
>> SomeClass someClass(ii++);
>> }
>>
>> {
>> SomeClass someClass2(ii++);
>> }
>>
>> {
>> SomeClass someClass3(ii++);
>> }
>> }
>> }
>> }
>>
>>
>> _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20160916/4d1f0e1d/attachment.html>