bug#31946: 27.0.50; The NSM should warn about more TLS problems

[Andy Moreton <andrewjmoreton@gmail.com>]

On 27/08/2019 8:23 am, Eli Zaretskii wrote:
>> From: Andy Moreton <andrewjmoreton@gmail.com>
>> Date: Mon, 26 Aug 2019 23:56:06 +0100
>> 
>> > Looks like something is wrong with how I'm using gnutls_free when it
>> > comes from the DLL.  I need to look into that.
>> 
>> Found it I think. gnutls_free is a pointer of type gnutls_free_function,
>> so the value returned from GetProcAddress is a data pointer, not a
>> function address. Something like the following patch appears to work
>> i.e. no longer crashes with the recipe above.
> 
> Yes, thanks.  I was struggling with the same thing, but evidently
> there's a limit to what I can do after 11PM at my age ;-)
> 
>> If this is the right fix then we will need to add e.g. DEF_DLL_DATA and
>> LOAD_DLL_DATA macros for runtime import of data (and an equivalent to
>> get_proc_addr to use the right types for runtime data imports).
> 
> I think there's no need to invent harness for a single user, we can
> simply do that by hand.  Please try the patch below, it's against the
> current master (it also removes the unnecessary intermediate buffer
> allocation while at that).

Your patch works (tested with package-list and gnus) - please install.

Thanks,

    AndyM