Re: [PATCH 2/4] msvcp90: Call exception ctor in logic_error and runtime_error ctor.

[Piotr Caban]

Hi Daniel,

this patch doesn't look right. The runtime_error class is supposed to 
override "what" virtual function.

I'm attaching a simple test case that shows that name field should not 
be set to passed string (at least in msvcp60).

Thanks,
Piotr

[PATCH 2/4] msvcp90: Call exception ctor in logic_error and runtime_error ctor.

[Daniel Lehman]

> I'm attaching a simple test case that shows that name field should not be set to passed string (at least in msvcp60).

Not setting name to the passed string breaks range_error for patch 4.  I'll take a closer look

Thanks
daniel

[PATCH 2/4] msvcp90: Call exception ctor in logic_error and runtime_error ctor.

[Piotr Caban]

On 10/13/17 18:14, Daniel Lehman wrote:
> > I'm attaching a simple test case that shows that name field should not be set to passed string (at least in msvcp60).
>
> Not setting name to the passed string breaks range_error for patch 4.  I'll take a closer look

This probably makes first patch incorrect. The range_error should 
probably use MSVCP_runtime_error_what in vtable. Also the class might 
have been changed in newer version of C-runtime (I wonder what's the 
value of sizeof(std::range_error) and sizeof(std::runtime_error)).

Thanks,
Piotr