November 2007 - Wine-Devel - list.winehq.org

winetest_interactive : how does it work?

by Paul Vriens

Hi, I just ran winetest (by double clicking on the executable) on my WinXP box. I was missing some advpack popups that were there before. Francois sent in a patch to only do some particular tests when winetest_interactive is true. I would assume that clicking on a executable is interactive, not? AFAIK winetest_interactive is set to true|1 if the WINETEST_INTERACTIVE variable is set to 1 (include/wine/test.h). In programs/winetest/main.c we however only set WINETEST_INTERACTIVE to 0: 692 interactive_no[] = "WINETEST_INTERACTIVE=0", shouldn't WINETEST_INTERACTIVE be set according to the interactive variable in that same file? -- Cheers, Paul.

Nov. 16, 2007

Re: server: Add a call for retrieving the image file name of a process.

by Alexandre Julliard

Robert Shearman <rob(a)codeweavers.com> writes: > --- > dlls/ntdll/process.c | 20 +++++++++++++++++++- > dlls/ntdll/tests/info.c | 36 ++++++++++++++++++++++++++++++++++++ > server/process.c | 31 +++++++++++++++++++++++++++++++ > server/protocol.def | 9 +++++++++ > 4 files changed, 95 insertions(+), 1 deletions(-) You should use the existing get_dll_info request for that. -- Alexandre Julliard julliard(a)winehq.org

Nov. 16, 2007

Re: ntdll: send_debug_event: don't send uninitialized bytes

by Robert Shearman

Dan Kegel wrote: > +/* Returns size of initialized part of a particular EXCEPTION_RECORD */ > +static int exception_record_size( const EXCEPTION_RECORD *rec ) > +{ > + return sizeof(*rec) > + - (EXCEPTION_MAXIMUM_PARAMETERS-rec->NumberParameters) * sizeof(ULONG_PTR); > +} FIELD_OFFSET exists for this exact purpose. -- Rob Shearman

Nov. 16, 2007

Re: [1/6] try 2: msvcrt: Implement strcpy_s

by Stefan Dösinger

Am Freitag, 16. November 2007 11:25:39 schrieb Stefan Dösinger: > > Did you test with invalid pointers (crash / handled graceful)? > > No, good idea, I'll try that now This crashes

Nov. 16, 2007

Re: [1/6] try 2: msvcrt: Implement strcpy_s

by Stefan Dösinger

Am Freitag, 16. November 2007 00:45:47 schrieb Detlef Riekenberg: > On Do, 2007-11-15 at 17:35 +0100, Stefan Dösinger wrote: > > + memset(dest, 'X', sizeof(dest)); > > + ret = pstrcpy_s(dest, sizeof(dest), small); > > + ok(ret == 0, "Copying a string into a big enough destination > > returned %d, expected 0\n", ret); > > + ok(dest[0] == 's' && dest[1] == 'm' && dest[2] == 'a' && dest[3] > > == 'l' && > > + dest[4] == 'l' && dest[5] == '\0'&& dest[6] == 'X' && dest[7] > > == 'X', > > Is it really needed to check the memory after the end of the string? > Does an application depend on this? It shows that the msdn documents the function incorrectly: It says the string is unmodified, which is apparently not the case. > Did you test with invalid pointers (crash / handled graceful)? No, good idea, I'll try that now

Nov. 16, 2007

re: dlls/msi/string.c -- fix use of signed versus unsigned (RESEND)

by Dan Kegel

Gerald Pfeifer wrote: > How do you propose to address the following? > string.c: In function 'msi_addstring': > string.c:208: warning: comparison between signed and unsigned For starters: diff --git a/configure.ac b/configure.ac index 3d62f45..c97aa8e 100644 --- a/configure.ac +++ b/configure.ac @@ -1134,6 +1134,7 @@ then WINE_TRY_CFLAGS([-fno-strict-aliasing]) WINE_TRY_CFLAGS([-Wdeclaration-after-statement]) WINE_TRY_CFLAGS([-Wwrite-strings]) + WINE_TRY_CFLAGS([-Wno-sign-compare]) dnl Check for noisy string.h saved_CFLAGS="$CFLAGS" Then, after we've fixed all the real bugs, maybe consider going back and turning on the signed/unsigned comparison warning.

Nov. 16, 2007

Re: [2/6] try 2: msvcrt: implement strcat_s

by Dmitry Timoshkov

"Stefan Dösinger" <stefan(a)codeweavers.com> wrote: > +@ cdecl strcat_s(str str) MSVCRT_strcat_s ... > +int CDECL MSVCRT_strcat_s( char* dst, MSVCRT_size_t elem, const char* src ) The spec file entry still has wrong number of parameters. -- Dmitry.

Nov. 16, 2007

Re: [3/6] try 2: msvcrt: implement _mbsnbcpy_s

by Detlef Riekenberg

On Do, 2007-11-15 at 17:37 +0100, Stefan Dösinger wrote: > + const unsigned char small[] = "small"; > + int ret; > + memset(dest, 'X', sizeof(dest)); > + ret = p_mbsnbcpy_s(dest, sizeof(dest), small, sizeof(small)); > + ok(ret == 0, "_mbsnbcpy_s: Copying a string into a big enough > destination returned %d, expected 0\n", ret); > + ok(dest[0] == 's' && dest[1] == 'm' && dest[2] == 'a' && dest[3] > == 'l' && > + dest[4] == 'l' && dest[5] == '\0'&& dest[6] == 'X' && dest[7] > == 'X you can use memcmp, when small is defined different: + const unsigned char small[] = "small\0XX"; I also suggest to make the buffer larger as you tell the api and always terminate the buffer unsigned char dest[10]; dest[sizeof(dest)-1] = '\0'; -- By by ... Detlef

Nov. 15, 2007

Re: [1/6] try 2: msvcrt: Implement strcpy_s

by Detlef Riekenberg

On Do, 2007-11-15 at 17:35 +0100, Stefan Dösinger wrote: > + memset(dest, 'X', sizeof(dest)); > + ret = pstrcpy_s(dest, sizeof(dest), small); > + ok(ret == 0, "Copying a string into a big enough destination > returned %d, expected 0\n", ret); > + ok(dest[0] == 's' && dest[1] == 'm' && dest[2] == 'a' && dest[3] > == 'l' && > + dest[4] == 'l' && dest[5] == '\0'&& dest[6] == 'X' && dest[7] > == 'X', Is it really needed to check the memory after the end of the string? Does an application depend on this? --- cut --- memset(dest, 'X', sizeof(dest)-1); /* never read past the buffer, when we check the result */ dest[sizeof(dest)-1] = '\0'; ret = pstrcpy_s(dest, sizeof(dest), small); ok( !ret && !lstrcmpA(dest, small), "got %d and '%s' (expected '0' and '%s')\n", ret, dest, small); -- cut -- For my tests with buffers, i found it useful to test with: - buffer present, but size is 0 - buffer one byte smaller as the string (no space left for the '\0') - buffer large enough for the string, but no space left for the '\0' - buffer exact as needed - buffer one byte larger as needed - empty source string - NULL source - NULL target This allow you to verify, how the terminating '\0' is handled. Did you test with invalid pointers (crash / handled graceful)? -- By by ... Detlef

Nov. 15, 2007

Re: [PATCH] bits: Stub version of qmgr with svchost and proxy (5/12)

by Detlef Riekenberg

On Do, 2007-11-15 at 08:22 -0800, Roy Shea wrote: > This adds svchost, proxying, and a stub version of the qmgr.dll > process. The > functional code in this patch is boilerplate required to run a DLL as > a > service via svchost. > --- > dlls/qmgr/Makefile.in | 6 ++- > dlls/qmgr/factory.c | 85 +++++++++++++++++++++++ > dlls/qmgr/qmgr.c | 8 ++- > dlls/qmgr/qmgr.inf | 32 +++++++++ > dlls/qmgr/qmgr.spec | 1 + > dlls/qmgr/qmgr_main.c | 98 +++++++++++++++++++++++++-- > dlls/qmgr/qmgr_private.h | 14 ++++- > dlls/qmgr/qmgr_service.c | 156 > +++++++++++++++++++++++++++++++++++++++++++ > dlls/qmgr/rsrc.rc | 21 ++++++ > dlls/qmgrprxy/Makefile.in | 25 +++++++ > dlls/qmgrprxy/qmgrprxy.idl | 20 ++++++ > dlls/qmgrprxy/qmgrprxy.spec | 5 ++ > 12 files changed, 461 insertions(+), 10 deletions(-) It would be better, when you can reduce the Patchsize, but I have no knowledge about the code. (must everything from factory.c, qmgr_main.c and qmgr_service.c in the same Patch?) qmgrprxy looks really splitable from the rest as a seperate Patch. > -0,0 +1,5 @@ > +@ stdcall -private DllCanUnloadNow() > +@ stdcall -private DllGetClassObject(ptr ptr ptr) > +@ stdcall -private DllRegisterServer() > +@ stdcall -private DllUnregisterServer() > +@ stdcall GetProxyDllInfo(ptr ptr) I'm unable to find the code for this functions. Is everything build from qmgrprxy.idl ? -- By by ... Detlef

Nov. 15, 2007