Is PureMalloc actually... pure?

[Rob Rau]

On Saturday, 11 January 2020 at 19:05:48 UTC, Chris Katko wrote:
> Pure implies (*requires) no side effects, right?
>
>>UNIX 98 requires that errno be set to ENOMEM upon failure. 
>>Purity is achieved by saving and restoring the value of errno, 
>>thus behaving as if it were never changed.
>
> Saving and restoring a global variable is is __not__ the same 
> as no side effects. without a mutex, two threads could easily 
> create a date race.
>
> void* pureMalloc()(size_t size) @trusted pure @nogc nothrow
> {
>     const errnosave = fakePureErrno;
>     void* ret = fakePureMalloc(size);
>     fakePureErrno = errnosave;
>     return ret;
> }
>
> I mean, on a practical level, malloc almost never fails so this 
> probably hasn't shown up as a bug in 99% of code.

To be posix compliant errno must be thread safe. In linux I 
believe this is done by each thread having its own thread local 
errno.

-Rob