From the discussion in <https://bugs.winehq.org/show_bug.cgi?id=50293>, these
patches don't seem to be providing any value (all they are doing is causing
arbitrarily chosen numbers to be shown in the dialog, which don't reflect
anything about actual functionality).
A possible benefit of these patches is anticipating their need by other
programs. However, (a) most of these DLLs are internal helpers used by other
(public) DirectX APIs, (b) the case of a missing DLL is usually quite easy to
debug anyway, (c) such a stance leaves us with the maintenance burden of far
more code that will never be necessary than code that actually turns out to be
useful.
Since (1) upstream has appeared reticent to accept them, (2) a proposal for
their removal in the bug report met with no apparent objection, (3) they have
already caused at least two regressions [however easy to fix], namely bug 50395
and bug 50700, (4) they present a noticeable maintenance burden, if a trivial
one, (5) as explained they provide no real benefit, remove them.
This was originally the third patch in a series, introduced by ec4e719cd6. The
first patch and the series itself was written by Mark Jansen and focused on an
incorrect handling of NULL in QueryActCtx(). It's not clear what those were
actually for, but they're upstream now.
My guess is that Sebastian noticed this "bug" while writing tests and wrote the
patch to fix it, as it's an easy one-liner. My guess is that he never submitted
it upstream either because it wasn't clearly correct (e.g. he wanted to actually
implement ACTIVATION_CONTEXT_SECTION_ASSEMBLY_INFORMATION) or because of
disagreements with upstream over project direction. As it happens there is a way
to get ACTIVATION_CONTEXT_SECTION_ASSEMBLY_INFORMATION to return valid data—pass
a NULL or empty string as the section name.
Anyway, the patch is likely to have never actually helped anything, and even if
it did, it's a one-liner with a FIXME message already attached, and so won't be
hard to debug again if necessary.
