Re: [PATCH] Allow new/delete operator deletion only for replaceable.

[Marc Glisse <marc.glisse@inria.fr>]

On Fri, 10 Apr 2020, Martin Liška wrote:

> On 4/9/20 10:13 AM, Jonathan Wakely wrote:
>> On Thu, 9 Apr 2020 at 09:05, Marc Glisse wrote:
>>> Note that the matching is not 1-to-1. Array vs non-array and
>>> aligned vs non-aligned seem important, but sized and unsized delete can
>>> both match the same new, IIUC.
>> 
>> Right.
>> 
>>> Not sure about the nothrow versions...
>> 
>> This is valid, and mixes the nothrow new with non-nothrow delete:
>> 
>> delete new (std::nothrow) int;
>> 
>
> All right, there's a patch candidate that comes up with the list of possible 
> pairs.
> For better readability, I present demangled names:
>
> $ cat /tmp/pairs.txt | c++filt
> 	  "operator new(unsigned long):operator delete(void*)" ,
> 	  "operator new(unsigned long):operator delete(void*, unsigned long)" 
> ,
> 	  "operator new(unsigned long):operator delete(void*, std::nothrow_t 
> const&)" ,
> 	  "operator new(unsigned long, std::nothrow_t const&):operator 
> delete(void*)" ,
> 	  "operator new(unsigned long, std::nothrow_t const&):operator 
> delete(void*, unsigned long)" ,
> 	  "operator new(unsigned long, std::nothrow_t const&):operator 
> delete(void*, std::nothrow_t const&)" ,
> 	  /* non-[] operators with alignment.  */
> 	  "operator new(unsigned long, std::align_val_t):operator 
> delete(void*, unsigned long, std::align_val_t)" ,
> 	  "operator new(unsigned long, std::align_val_t):operator 
> delete(void*, std::align_val_t)" ,
> 	  "operator new(unsigned long, std::align_val_t):operator 
> delete(void*, std::align_val_t, std::nothrow_t const&)" ,
> 	  "operator new(unsigned long, std::align_val_t, std::nothrow_t 
> const&):operator delete(void*, unsigned long, std::align_val_t)" ,
> 	  "operator new(unsigned long, std::align_val_t, std::nothrow_t 
> const&):operator delete(void*, std::align_val_t)" ,
> 	  "operator new(unsigned long, std::align_val_t, std::nothrow_t 
> const&):operator delete(void*, std::align_val_t, std::nothrow_t const&)" ,
> 	  /* [] operators.  */
> 	   "operator new[](unsigned long):operator delete[](void*)" ,
> 	   "operator new[](unsigned long):operator delete[](void*, unsigned 
> long)" ,
> 	   "operator new[](unsigned long):operator delete[](void*, 
> std::nothrow_t const&)" ,
> 	   "operator new[](unsigned long, std::nothrow_t const&):operator 
> delete[](void*)" ,
> 	   "operator new[](unsigned long, std::nothrow_t const&):operator 
> delete[](void*, unsigned long)" ,
> 	   "operator new[](unsigned long, std::nothrow_t const&):operator 
> delete[](void*, std::nothrow_t const&)" ,
> 	  /* [] operators with alignment.  */
> 	   "operator new[](unsigned long, std::align_val_t):operator 
> delete[](void*, unsigned long, std::align_val_t)" ,
> 	   "operator new[](unsigned long, std::align_val_t):operator 
> delete[](void*, std::align_val_t)" ,
> 	   "operator new[](unsigned long, std::align_val_t):operator 
> delete[](void*, std::align_val_t, std::nothrow_t const&)" ,
> 	   "operator new[](unsigned long, std::align_val_t, std::nothrow_t 
> const&):operator delete[](void*, unsigned long, std::align_val_t)",
> 	   "operator new[](unsigned long, std::align_val_t, std::nothrow_t 
> const&):operator delete[](void*, std::align_val_t)",
> 	   "operator new[](unsigned long, std::align_val_t, std::nothrow_t 
> const&):operator delete[](void*, std::align_val_t, std::nothrow_t const&)",
>
> Marc pointed out that some targets do not use the leading '_' (or use 2 
> dashes?) for mangled named?
> Patch can bootstrap on x86_64-linux-gnu and survives regression tests.

How do you handle platforms where size_t is not unsigned long?

-- 
Marc Glisse