[1/2] richedit: Created initial tests for windowless richedit controls. (try 2)
kumbayo84 at arcor.de
Sat Nov 8 09:40:29 CST 2008
On Samstag 08 November 2008, Dylan Smith wrote:
> (Ensured compiler independence by modifying byte code templates)
> The initial tests added simply call CreateTextServices, and query for
> the ITextServices interface, then cleans up.
> Since CreateTextServices needs an implementation of ITextHost to be
> given, this patch needed a stub implementation. This required wrappers
> for converting thiscall to stdcall which are now implementing by
> adapting a byte code wrapper template. Although byte code isn't portable
> across processors, this is only needed for x86 platforms so it is only
> meant to be compiler independent.
> In order to investigate what is happening in the tests, it is helpful
> to trace the calls to ITextHost methods by the richedit controls. I
> realized it would be too verbose to trace all these calls by default,
> so I wrapped them all with a check for winedebug > 1, since normally
> winedebug is 0 or 1.
> dlls/riched20/tests/Makefile.in | 3 +-
> dlls/riched20/tests/txtsrv.c | 593 +++++++++++++++++++++++++++++++++++++++
> 2 files changed, 595 insertions(+), 1 deletions(-)
> create mode 100644 dlls/riched20/tests/txtsrv.c
This page  seems to suggest memory returned from HeapAlloc
is marked as not executable.
Other tests (ntdll exception) are using VirtualAlloc()
for dynamically created code.
> +/* The following x86 code string converts between the thiscall and stdcall
> + * calling convetions. The thiscall calling convention places the This
> + * pointer in ecx on the x86 platform, and the stdcall calling convention
> + * pushes the This pointer on the stack as the first argument.
> + *
> + * The wrapper's code finishes by jumping to the real function.
> + *
> + * Byte codes are used so that a copy of it can be modified to use
> + * for each method in ITextHost. */
> -/* The following x86 code string converts between the thiscall and stdcall
> +/* The following x86 code strings convert between the thiscall and stdcall
> * calling convetions. The thiscall calling convention places the This
Here you introduce comments that you seem to change in the second patch.
Maybe use the right text in the first patch?
Additionally the term code string sounds odd.
Maybe replace that with "The following x86 assembler code..."
or something similar?
> #define ITextServices_QueryInterface(p,a,b) (p)->lpVtbl->QueryInterface(p,a,b)
> #define ITextServices_AddRef(p) (p)->lpVtbl->AddRef(p)
> #define ITextServices_Release(p) (p)->lpVtbl->Release(p)
> -/*** ITextServices methods ***/
> -#define ITextServices_TxSendMessage(p,a,b,c,d) (p)->lpVtbl->TxSendMessage(p,a,b,c,d)
> -#define ITextServices_TxDraw(p,a,b,c,d,e,f,g,h,i,j,k,l) (p)->lpVtbl->TxDraw(p,a,b,c,d,e,f,g,h,i,j,k,l)
Why do you only move some of those #defines part of ITextServices
and not the one part of IUnknown ?
I do not know much about those #defines but this seems odd
Maybe i am missing something but why do you need those thiscall wrappers?
Wouldn't it be possible to use
the DEFINE_THISCALL_WRAPPER and THISCALL macros like in txtserv.c?
More information about the wine-devel