Re: ntdll: Fix the file mapping view leak in the case of module loading failure. Take 2.

[Alexandre Julliard]

Dmitry Timoshkov <dmitry(a)baikal.ru> writes:
> @@ -1548,6 +1549,11 @@ static NTSTATUS load_native_dll( LPCWSTR load_path, LPCWSTR name, HANDLE file,
>      *pwm = wm;
>      status = STATUS_SUCCESS;
>  done:
> +    if (status)
> +    {
> +        if (wm) free_modref( wm );
> +        else if (module) NtUnmapViewOfSection( NtCurrentProcess(), module );
> +    }

I don't see any evidence that you are addressing the issues that caused
us to not do that in the first place.

-- 
Alexandre Julliard
julliard(a)winehq.org

ntdll: Fix the file mapping view leak in the case of module loading failure. Take 2.

[Dmitry Timoshkov]

Alexandre Julliard <julliard(a)winehq.org> wrote:
> I don't see any evidence that you are addressing the issues that caused
> us to not do that in the first place.

Because the only thing I want to see fixed at this point is a view leak,
it's still better than current state of things.

Without this fix Wine shows exactly the same broken behaviour as NT4 does in
the loader test when LoadLibrary is called on an already mapped image:
dangling/leaked file handle, inability to overwrite/delete an executable file.

-- 
Dmitry.

ntdll: Fix the file mapping view leak in the case of module loading failure. Take 2.

[Alexandre Julliard]

Dmitry Timoshkov <dmitry(a)baikal.ru> writes:
> Alexandre Julliard <julliard(a)winehq.org> wrote:
> > I don't see any evidence that you are addressing the issues that caused
> > us to not do that in the first place.
>
> Because the only thing I want to see fixed at this point is a view leak,
> it's still better than current state of things.

I don't think that deleting an object that may have references is making
things better.

-- 
Alexandre Julliard
julliard(a)winehq.org