[PATCH] ntdll: Fix unwinding from leaf function on x86_64.

[Derek Lesho]

---
 dlls/ntdll/signal_x86_64.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/dlls/ntdll/signal_x86_64.c b/dlls/ntdll/signal_x86_64.c
index 54871b8302..21e7098876 100644
--- a/dlls/ntdll/signal_x86_64.c
+++ b/dlls/ntdll/signal_x86_64.c
@@ -4232,7 +4232,7 @@ void WINAPI RtlUnwindEx( PVOID end_frame, PVOID target_ip, EXCEPTION_RECORD *rec

new_context.Rip = *(ULONG64 *)context->Rsp;
         new_context.Rsp = context->Rsp + sizeof(ULONG64);
-        dispatch.EstablisherFrame = new_context.Rsp;
+        dispatch.EstablisherFrame = context->Rsp;
         dispatch.LanguageHandler = NULL;

unwind_done:

[Marvin]

Hi,

While running your changed tests, I think I found new failures.
Being a bot and all I'm not very good at pattern recognition, so I might be
wrong, but could you please double-check?

Full results can be found at:
https://testbot.winehq.org/JobDetails.pl?Key=49463

Your paranoid android.

=== debian9b (64 bit WoW report) ===

ntdll:
env.c:498: Test failed: wrong end ptr 0x1238c/0x12390

[Derek Lesho]

I'm not sure why that test fails, that test shouldn't even trigger the
code I change.  By the way, I made a test case that compiles in MSVC.
It's slightly modified from the MSDN's seh example to showcase the
issue with leaf functions: https://godbolt.org/z/mo1-HV

On Fri, Mar 15, 2019 at 3:40 PM Marvin  wrote:
> Hi,
>
> While running your changed tests, I think I found new failures.
> Being a bot and all I'm not very good at pattern recognition, so I might be
> wrong, but could you please double-check?
>
> Full results can be found at:
> https://testbot.winehq.org/JobDetails.pl?Key=49463
>
> Your paranoid android.
>
> === debian9b (64 bit WoW report) ===
>
> ntdll:
> env.c:498: Test failed: wrong end ptr 0x1238c/0x12390

[Ken Thomases]

If no existing Wine test covers this case, then you should add a test both demonstrating that your change is correct and to prevent any future changes from accidentally re-breaking things.

-Ken
> On Mar 15, 2019, at 2:55 PM, Derek Lesho  wrote:
>
> I'm not sure why that test fails, that test shouldn't even trigger the
> code I change.  By the way, I made a test case that compiles in MSVC.
> It's slightly modified from the MSDN's seh example to showcase the
> issue with leaf functions: https://godbolt.org/z/mo1-HV
>
> On Fri, Mar 15, 2019 at 3:40 PM Marvin  wrote:
> > Hi,
> >
> > While running your changed tests, I think I found new failures.
> > Being a bot and all I'm not very good at pattern recognition, so I might be
> > wrong, but could you please double-check?
> >
> > Full results can be found at:
> > https://testbot.winehq.org/JobDetails.pl?Key=49463
> >
> > Your paranoid android.
> >
> > === debian9b (64 bit WoW report) ===
> >
> > ntdll:
> > env.c:498: Test failed: wrong end ptr 0x1238c/0x12390

[Derek Lesho]

Ken, I have written a "test-case" using MSVC, but I don't think I know
enough to write a test-case in wine, due to the manual nature of
setting up the assembly and the exception data.  By the way, the
compiler-explorer program linked above was not the correct one, this
is: https://godbolt.org/z/WYcNrE

But if you compile this in debug mode and run it in current wine, it
will crash, if you run it with either my patch or on windows 7, it
will succeed.

On Fri, Mar 15, 2019 at 4:06 PM Ken Thomases  wrote:
> If no existing Wine test covers this case, then you should add a test both demonstrating that your change is correct and to prevent any future changes from accidentally re-breaking things.
>
> -Ken
> > On Mar 15, 2019, at 2:55 PM, Derek Lesho  wrote:
> >
> > I'm not sure why that test fails, that test shouldn't even trigger the
> > code I change.  By the way, I made a test case that compiles in MSVC.
> > It's slightly modified from the MSDN's seh example to showcase the
> > issue with leaf functions: https://godbolt.org/z/mo1-HV
> >
> > On Fri, Mar 15, 2019 at 3:40 PM Marvin  wrote:
> > > Hi,
> > >
> > > While running your changed tests, I think I found new failures.
> > > Being a bot and all I'm not very good at pattern recognition, so I might be
> > > wrong, but could you please double-check?
> > >
> > > Full results can be found at:
> > > https://testbot.winehq.org/JobDetails.pl?Key=49463
> > >
> > > Your paranoid android.
> > >
> > > === debian9b (64 bit WoW report) ===
> > >
> > > ntdll:
> > > env.c:498: Test failed: wrong end ptr 0x1238c/0x12390

[Derek Lesho]

Signed-off-by: Derek Lesho

On Fri, Mar 15, 2019 at 3:28 PM Derek Lesho  wrote:
> dlls/ntdll/signal_x86_64.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/dlls/ntdll/signal_x86_64.c b/dlls/ntdll/signal_x86_64.c
> index 54871b8302..21e7098876 100644
> --- a/dlls/ntdll/signal_x86_64.c
> +++ b/dlls/ntdll/signal_x86_64.c
> @@ -4232,7 +4232,7 @@ void WINAPI RtlUnwindEx( PVOID end_frame, PVOID target_ip, EXCEPTION_RECORD *rec
>      new_context.Rip = *(ULONG64 *)context->Rsp;
>      new_context.Rsp = context->Rsp + sizeof(ULONG64);
> -
>    dispatch.EstablisherFrame = new_context.Rsp;
> -
>    dispatch.EstablisherFrame = context->Rsp;
>    dispatch.LanguageHandler = NULL;
>
> unwind_done:
>
> --
> 2.20.1