Re: dlls/rsaenh/rsaenh.c

[Alexandre Julliard]

Michael Jung  writes:
> Changelog:
>  CPAcquireContext: Set last error to ERROR_SUCCESS in case of success.

There are very few functions that set last error on success, do you
have a test demonstrating that Windows does it in this case?

dlls/rsaenh/rsaenh.c

[Michael Jung]

On Tuesday 09 November 2004 20:21, Alexandre Julliard wrote:
> Michael Jung  writes:
> > Changelog:
> >  CPAcquireContext: Set last error to ERROR_SUCCESS in case of success.
>
> There are very few functions that set last error on success, do you
> have a test demonstrating that Windows does it in this case?

The test is in dlls/advapi32/tests/crypt.c, line 117. It passes at least on 
WinXP SP2. What I don't know is if this should be set in 
advapi32/CryptAcquireContext or in rsaenh/CPAcquireContext. However, setting 
it in rsaenh can do no harm.

Greetings,
Michael

dlls/rsaenh/rsaenh.c

[Alexandre Julliard]

Michael Jung  writes:
> The test is in dlls/advapi32/tests/crypt.c, line 117. It passes at least on 
> WinXP SP2.

OK great, I missed that one.
> What I don't know is if this should be set in 
> advapi32/CryptAcquireContext or in rsaenh/CPAcquireContext. However, setting 
> it in rsaenh can do no harm.

Famous last words... It definitely matters where it's done, because we
can use various native/builtin combinations, so we have to do it where
Windows does.

dlls/rsaenh/rsaenh.c

[Michael Jung]

On Tuesday 09 November 2004 20:52, Alexandre Julliard wrote:
> Michael Jung  writes:
> > The test is in dlls/advapi32/tests/crypt.c, line 117. It passes at least
> > on WinXP SP2.
>
> OK great, I missed that one.
> > What I don't know is if this should be set in
> > advapi32/CryptAcquireContext or in rsaenh/CPAcquireContext. However,
> > setting it in rsaenh can do no harm.
>
> Famous last words... It definitely matters where it's done, because we
> can use various native/builtin combinations, so we have to do it where
> Windows does.

I agree that if this is done on Windows in advapi32, we have to do it there, 
too. What I meant to say was the following:

There are basically two possibilities:
1.) It's done in rsaenh. Then we are ok.
2.) It's done in advapi32: Then it doesn't matter if we already have set it in 
rsaenh, because it will be overwritten.

Greetings,
Michael

dlls/rsaenh/rsaenh.c

[Alexandre Julliard]

Michael Jung  writes:
> I agree that if this is done on Windows in advapi32, we have to do it there, 
> too. What I meant to say was the following:
>
> There are basically two possibilities:
> 1.) It's done in rsaenh. Then we are ok.
> 2.) It's done in advapi32: Then it doesn't matter if we already have set it in 
> rsaenh, because it will be overwritten.

But if Windows doesn't set it in rsaenh, then we are just hiding a
bug. And of course there's always the possibility that native advapi32
somehow depends on rsaenh not doing it; I know it's unlikely, but the
point is we can't just do things differently and assume that it will
work.

dlls/rsaenh/rsaenh.c

[Michael Jung]

On Tuesday 09 November 2004 21:24, Alexandre Julliard wrote:
> Michael Jung  writes:
> > I agree that if this is done on Windows in advapi32, we have to do it
> > there, too. What I meant to say was the following:
> >
> > There are basically two possibilities:
> > 1.) It's done in rsaenh. Then we are ok.
> > 2.) It's done in advapi32: Then it doesn't matter if we already have set
> > it in rsaenh, because it will be overwritten.
>
> But if Windows doesn't set it in rsaenh, then we are just hiding a
> bug. And of course there's always the possibility that native advapi32
> somehow depends on rsaenh not doing it; I know it's unlikely, but the
> point is we can't just do things differently and assume that it will
> work.

The attached program produces the following output on my WinXP SP2:

Loaded rsaenh.dll successfully!
Got CPAcquireContext address!
Success! Last error: 00000000

So it seems that rsaenh is setting last error to ERROR_SUCCESS. If you wan't 
to try the program, make sure that there is a default key container named 
with your login name on your system. Otherwise CPAcquireContext will fail.

Greetings,
Michael

dlls/rsaenh/rsaenh.c

[Alexandre Julliard]

Michael Jung  writes:
> The attached program produces the following output on my WinXP SP2:
>
> Loaded rsaenh.dll successfully!
> Got CPAcquireContext address!
> Success! Last error: 00000000

Cool, thanks for taking the time to do this. I'll put the patch in.