On 09/09/2018 11:16 AM, Roger Zoellner wrote:
> +static DWORD WINAPI thread_main(LPVOID lpParam)
> +{
> +    /* Thread is created with CREATE_SUSPENDED flag set and isn't doing anything */
> +}

You should still probably do

(void)lpParam;
return 0;

to avoid compiler warnings about unused parameters and lacking a return
value.

> +    /* Test required affinity mask, all bits set */
> +    req_aff = ~0u;

This isn't all bits set for 64-bit targets. Try

req_aff = ~(ULONG_PTR)0;

for that.

> +    status = pNtSetInformationThread(thread, class, &req_aff, req_aff_length);
> +    ok(status == STATUS_SUCCESS, "NtSetInformationThread failed. Expected STATUS_SUCCESS, got %08x (class=ThreadAffinityMask, case 2, req_aff=%08x, process_aff=%08x)\n", status, req_aff, process_aff);
> ...
> +    /* Test required affinity mask = process affinity mask with a prepended bit set */
> +    req_aff = (process_aff << 1) | process_aff;
> +    status = pNtSetInformationThread(thread, class, &req_aff, req_aff_length);
> +    ok(status == STATUS_SUCCESS, "NtSetInformationThread failed. Expected STATUS_SUCCESS, got %08x (class=ThreadAffinityMask, case 4, req_aff=%08x, process_aff=%08x)\n", status, req_aff, process_aff);

Might it be worth checking to see what's returned from
NtQueryInformationThread after trying to set extraneous mask bits?

Also, the tests that currently fail under Wine should be marked with
todo_wine until a proper fix is made, and such a fix would remove the
rodo_wines (unless the tests are part of a series that come after the fix).