October 2017 - Wine-Devel - list.winehq.org

Re: oleaut32: Add ARM support to DispCallFunc().
by André Hentschel Oct. 23, 2017

Oct. 23, 2017

Hi Donna and Welcome to the Wine Project! Whats your motivation for this patch? You did a very good job for a firsttimer! Some more notes below: Am 22.10.2017 um 03:04 schrieb Donna Whisnant: > Adds ARM ABI support to DispCallFunc() to allow IDispatch invoke calls > to succeed on ARM platforms. This change specifically targets only > 32-bit little-endian (ARMEL) platform CPUs. It's believed to likely be > compatible with 64-bit and/or big-endian (ARMEB) platforms, but testing > for those other platforms should be completed before enabling. > > Tested on Raspbian Stretch 2017-09-07 RPi image using a Qemu systemd > container to compile and run it on an x86-64 Host.. When running on > Raspberry Pi hardware, a 3G/1G split kernel needs to be used for Wine > to function. > > Signed-off-by: Donna Whisnant <dewhisna(a)dewtronics.com> > --- > dlls/oleaut32/typelib.c | 221 +++++++++++++++++++++++++++++++++++++++++++++++- > 1 file changed, 219 insertions(+), 2 deletions(-) > > diff --git a/dlls/oleaut32/typelib.c b/dlls/oleaut32/typelib.c > index ebf6d85..d7ef212 100644 > --- a/dlls/oleaut32/typelib.c > +++ b/dlls/oleaut32/typelib.c > @@ -6302,8 +6302,48 @@ static HRESULT WINAPI ITypeInfo_fnGetIDsOfNames( ITypeInfo2 *iface, > return DISP_E_UNKNOWNNAME; > } > > +#if defined(__arm__) && defined(__ARM_32BIT_STATE) && !defined(__ARMEB__) > +/* Note: this may work OK on 64-bit and/or ARMEB, but has currently only been tested for 32-bit ARMEL */ While i doubt there is something risky here regarding endianess you can keep the ARMEB check, but get rid of the 32bit check, it might be about Thunk(16bit) vs. ARM(32bit) but I'm not sure. Anyway, this won't work on 64bit, because the instruction set is totally different, but it inspires me to do the 64bit version as soon this lands upstream ;) > > -#ifdef __i386__ > +extern LONGLONG call_method( void *func, int nb_stk_args, const DWORD *stk_args, const DWORD *reg_args ); > +__ASM_GLOBAL_FUNC( call_method, > + /* r0 = *func > + * r1 = nb_stk_args > + * r2 = *stk_args (pointer to 'nb_stk_args' DWORD values to push on stack) > + * r3 = *reg_args (pointer to 8, 64-bit d0-d7 (double) values OR as 16, 32-bit s0-s15 (float) values, followed by 4, 32-bit (DWORD) r0-r3 values) > + */ > + > + "push {fp, lr}\n\t" /* Save frame pointer and return address (stack still aligned to 8 bytes) */ > + __ASM_CFI(".cfi_adjust_cfa_offset 4\n\t") > + __ASM_CFI(".cfi_rel_offset %fp,0\n\t") > + "add fp, sp, #4\n\t" /* Setup frame pointer */ Is this adjusted FP used somewhere? (I forgot a lot of things regarding arm32...) > + __ASM_CFI(".cfi_def_cfa_register %fp\n\t") > + > + "lsls r1, r1, #2\n\t" /* r1 = nb_stk_args * sizeof(DWORD) */ > + "beq 1f\n\t" /* Skip allocation if no stack args */ > + "add r2, r2, r1\n" /* Calculate ending address of incoming stack data */ > + "2:\tldr ip, [r2, #-4]!\n\t" /* Get next value */ > + "str ip, [sp, #-4]!\n\t" /* Push it on the stack */ > + "subs r1, r1, #4\n\t" /* Decrement count */ > + "bgt 2b\n\t" /* Loop till done */ > + > + "1:\tvldm r3!, {s0-s15}\n\t" /* Load the s0-s15/d0-d7 arguments */ > + "mov ip, r0\n\t" /* Save the function call address to ip before we nuke r0 with arguments to pass */ > + "ldm r3, {r0-r3}\n\t" /* Load the r0-r3 arguments */ > + > + "blx ip\n\t" /* Call the target function */ > + > + "sub sp, fp, #4\n\t" /* Clean the stack using 'fp' to allow for either stdcall or cdecl calls */ > + __ASM_CFI(".cfi_def_cfa %sp,4\n\t") > + "pop {fp, pc}\n\t" /* Restore FP and return */ > + __ASM_CFI(".cfi_same_value %fp\n\t") /* Here for completeness and symmetry with other cpus, but won't be reached */ > + ) > + > +/* same function but returning single/double floating point */ > +static float (* const call_float_method)(void *, int, const DWORD *, const DWORD *) = (void *)call_method; > +static double (* const call_double_method)(void *, int, const DWORD *, const DWORD *) = (void *)call_method; > + > +#elif defined(__i386__) > > extern LONGLONG call_method( void *func, int nb_args, const DWORD *args, int *stack_offset ); > __ASM_GLOBAL_FUNC( call_method, > @@ -6637,7 +6677,184 @@ DispCallFunc( > void* pvInstance, ULONG_PTR oVft, CALLCONV cc, VARTYPE vtReturn, UINT cActuals, > VARTYPE* prgvt, VARIANTARG** prgpvarg, VARIANT* pvargResult) > { > -#ifdef __i386__ > +#if defined(__arm__) && defined(__ARM_32BIT_STATE) && !defined(__ARMEB__) > +/* Note: this may work OK on 64-bit and/or ARMEB, but has currently only been tested for 32-bit ARMEL */ Same as above > + int argspos; > + void *func; > + UINT i; > + DWORD *args; > + struct { > + union { > + float s[16]; > + double d[8]; > + } sd; > + DWORD r[4]; > + } regs; > + int rcount; /* 32-bit register index count */ > + int scount; /* single-precision float register index count (will be incremented twice for doubles, plus alignment) */ > + > + TRACE("(%p, %ld, %d, %d, %d, %p, %p, %p (vt=%d))\n", > + pvInstance, oVft, cc, vtReturn, cActuals, prgvt, prgpvarg, > + pvargResult, V_VT(pvargResult)); You can go up to 100 chars per line, so be free to rearrange this > + > + if (cc != CC_STDCALL && cc != CC_CDECL) > + { > + FIXME("unsupported calling convention %d\n",cc); > + return E_INVALIDARG; > + } > + > + argspos = 0; > + rcount = 0; > + scount = 0; > + > + /* Determine if we need to pass a pointer for the return value as arg 0. If so, do that */ > + /* first as it will need to be in the 'r' registers: */ > + switch (vtReturn) > + { > + case VT_DECIMAL: > + case VT_VARIANT: > + regs.r[rcount++] = (DWORD)pvargResult; /* arg 0 is a pointer to the result */ > + break; > + case VT_HRESULT: > + WARN("invalid return type %u\n", vtReturn); > + return E_INVALIDARG; > + default: /* And all others are in 'r', 's', or 'd' registers or have no return value */ > + break; > + } > + > + if (pvInstance) > + { > + const FARPROC *vtable = *(FARPROC **)pvInstance; > + func = vtable[oVft/sizeof(void *)]; > + regs.r[rcount++] = (DWORD)pvInstance; /* the This pointer is always the first parameter */ > + } > + else func = (void *)oVft; > + > + /* maximum size for an argument is sizeof(VARIANT). Also allow for return pointer and stack alignment. */ > + args = heap_alloc(sizeof(VARIANT) * cActuals + sizeof(DWORD) * 4 ); > + > + for (i = 0; i < cActuals; i++) > + { > + VARIANT *arg = prgpvarg[i]; > + DWORD *pdwarg = (DWORD *)(arg); /* a reinterpret_cast of the variant, used for copying structures when they are split between registers and stack */ > + int ntemp; /* Used for counting words split between registers and stack */ > + > + switch (prgvt[i]) > + { > + case VT_EMPTY: > + break; > + case VT_R4: /* these must be 4-byte aligned, and put in 's' regs or stack, as they are single-floats */ > + if (scount < 16) > + { > + regs.sd.s[scount++] = V_R4(arg); > + } else { > + args[argspos++] = V_UI4(arg); > + } I think no curly brackets are necessary here at all, but we usually don't open them in the same line as the if/else statement for new code > + break; > + case VT_R8: /* these must be 8-byte aligned, and put in 'd' regs or stack, as they are double-floats */ > + case VT_DATE: > + if (scount < 15) > + { > + scount += (scount % 2); /* align scount to next whole double */ > + regs.sd.d[scount/2] = V_R8(arg); > + scount += 2; > + } else { > + scount = 16; /* Make sure we flag that all 's' regs are full */ > + argspos += (argspos % 2); /* align argspos to 8-bytes */ > + memcpy( &args[argspos], &V_R8(arg), sizeof(V_R8(arg)) ); > + argspos += sizeof(V_R8(arg)) / sizeof(DWORD); > + } > + break; > + case VT_I8: /* these must be 8-byte aligned, and put in 'r' regs or stack, as they are long-longs */ > + case VT_UI8: > + case VT_CY: > + if (rcount < 3) > + { > + rcount += (rcount % 2); /* align rcount to 8-byte register pair */ > + memcpy( &regs.r[rcount], &V_UI8(arg), sizeof(V_UI8(arg)) ); > + rcount += sizeof(V_UI8(arg)) / sizeof(DWORD); > + } else { > + rcount = 4; /* Make sure we flag that all 'r' regs are full */ > + argspos += (argspos % 2); /* align argspos to 8-bytes */ > + memcpy( &args[argspos], &V_UI8(arg), sizeof(V_UI8(arg)) ); > + argspos += sizeof(V_UI8(arg)) / sizeof(DWORD); > + } > + break; > + case VT_DECIMAL: /* these structures are 8-byte aligned, and put in 'r' regs or stack, can be split between the two */ > + case VT_VARIANT: > + /* 8-byte align 'r' and/or stack: */ > + if (rcount < 3) > + { > + rcount += (rcount % 2); > + } else { > + rcount = 4; > + argspos += (argspos % 2); > + } > + ntemp = sizeof(*arg) / sizeof(DWORD); > + while (ntemp > 0) > + { > + if (rcount < 4) > + { > + regs.r[rcount++] = *pdwarg++; > + } else { > + args[argspos++] = *pdwarg++; > + } > + --ntemp; > + } > + break; > + case VT_BOOL: /* VT_BOOL is 16-bit but BOOL is 32-bit, needs to be extended */ > + if (rcount < 4) > + { > + regs.r[rcount++] = V_BOOL(arg); > + } else { > + args[argspos++] = V_BOOL(arg); > + } > + break; > + default: > + if (rcount < 4) > + { > + regs.r[rcount++] = V_UI4(arg); > + } else { > + args[argspos++] = V_UI4(arg); > + } > + break; > + } > + TRACE("arg %u: type %s %s\n", i, debugstr_vt(prgvt[i]), debugstr_variant(arg)); > + } > + > + argspos += (argspos % 2); /* Make sure stack function alignment is 8-byte */ > + > + TRACE("rcount: %d, scount: %d, argspos: %d\n", rcount, scount, argspos); > + > + switch (vtReturn) > + { > + case VT_EMPTY: /* EMPTY = no return value */ > + case VT_DECIMAL: /* DECIMAL and VARIANT already have a pointer argument passed (see above) */ > + case VT_VARIANT: > + call_method( func, argspos, args, (DWORD*)&regs ); > + break; > + case VT_R4: > + V_R4(pvargResult) = call_float_method( func, argspos, args, (DWORD*)&regs ); > + break; > + case VT_R8: > + case VT_DATE: > + V_R8(pvargResult) = call_double_method( func, argspos, args, (DWORD*)&regs ); > + break; > + case VT_I8: > + case VT_UI8: > + case VT_CY: > + V_UI8(pvargResult) = call_method(func, argspos, args, (DWORD*)&regs ); > + break; > + default: > + V_UI4(pvargResult) = call_method( func, argspos, args, (DWORD*)&regs ); > + break; > + } > + heap_free( args ); > + if (vtReturn != VT_VARIANT) V_VT(pvargResult) = vtReturn; > + TRACE("retval: %s\n", debugstr_variant(pvargResult)); > + return S_OK; > + > +#elif defined(__i386__) > int argspos, stack_offset; > void *func; > UINT i; >

1 0

Re: [3/4] include: Add twain 2.x header for use in data sources.
by Henri Verbeet Oct. 23, 2017

Oct. 23, 2017

On 20 October 2017 at 21:16, Vincent Povirk <vincent(a)codeweavers.com> wrote: > diff --git a/include/twain2.h b/include/twain2.h > new file mode 100644 > index 0000000..8455649 > --- /dev/null > +++ b/include/twain2.h > @@ -0,0 +1,2131 @@ > +/* ======================================================================== *\ > + > + Copyright (C) 2007 TWAIN Working Group: Adobe Systems Incorporated, > + AnyDoc Software Inc., Eastman Kodak Company, Fujitsu Computer Products > + of America, JFL Peripheral Solutions Inc., Ricoh Corporation, and > + Xerox Corporation. All rights reserved. > + > + Copyright (C) 1991, 1992 TWAIN Working Group: Aldus, Caere, Eastman-Kodak, > + Hewlett-Packard and Logitech Corporations. All rights reserved. > + > + Copyright (C) 1997 TWAIN Working Group: Bell+Howell, Canon, DocuMagix, > + Fujitsu, Genoa Technology, Hewlett-Packard, Kofax Imaging Products, and > + Ricoh Corporation. All rights reserved. > + > + Copyright (C) 1998 TWAIN Working Group: Adobe Systems Incorporated, > + Canon Information Systems, Eastman Kodak Company, > + Fujitsu Computer Products of America, Genoa Technology, > + Hewlett-Packard Company, Intel Corporation, Kofax Image Products, > + JFL Peripheral Solutions Inc., Ricoh Corporation, and Xerox Corporation. > + All rights reserved. > + > + Copyright (C) 2000 TWAIN Working Group: Adobe Systems Incorporated, > + Canon Information Systems, Digimarc Corporation, Eastman Kodak Company, > + Fujitsu Computer Products of America, Hewlett-Packard Company, > + JFL Peripheral Solutions Inc., Ricoh Corporation, and Xerox Corporation. > + All rights reserved. > + Doesn't the above essentially say we can't distribute this?

3 5

Re: [v2 PATCH 1/2] gdiplus/metafile: Implement EmfPlusRecordTypeSetClipRegion playback
by Vincent Povirk Oct. 23, 2017

Oct. 23, 2017

Is there a reason to clone the region instead of using it directly from the object table?

1 1

Re: [PATCH] ddraw/tests: Avoid dereferencing a null pointer in case of failure.
by Henri Verbeet Oct. 23, 2017

Oct. 23, 2017

On 22 October 2017 at 00:32, Zebediah Figura <z.figura12(a)gmail.com> wrote: > This should fix a persistent crash on one of the XP test machines. > > Signed-off-by: Zebediah Figura <z.figura12(a)gmail.com> > --- Why does it fail? It shouldn't. Does this run against Microsoft's ddraw, or against some VM replacement?

2 2

Re: [PATCH] msvcrt: Implement _memicmp_l()
by Marvin Oct. 23, 2017

Oct. 23, 2017

Hi, While running your changed tests on Windows, 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=33714 Your paranoid android. === wvistau64 (32 bit string) === string.c:178: Test failed: errno = 0, expected EINVAL string.c:185: Test failed: errno = 0, expected EINVAL string.c:192: Test failed: errno = 0, expected EINVAL string.c:221: Test failed: errno = 0, expected EINVAL string.c:228: Test failed: errno = 0, expected EINVAL string.c:235: Test failed: errno = 0, expected EINVAL === w7u (32 bit string) === string.c:178: Test failed: errno = 0, expected EINVAL string.c:185: Test failed: errno = 0, expected EINVAL string.c:192: Test failed: errno = 0, expected EINVAL string.c:221: Test failed: errno = 0, expected EINVAL string.c:228: Test failed: errno = 0, expected EINVAL string.c:235: Test failed: errno = 0, expected EINVAL === w1064 (32 bit string) === string.c:178: Test failed: errno = 0, expected EINVAL string.c:185: Test failed: errno = 0, expected EINVAL string.c:192: Test failed: errno = 0, expected EINVAL string.c:221: Test failed: errno = 0, expected EINVAL string.c:228: Test failed: errno = 0, expected EINVAL string.c:235: Test failed: errno = 0, expected EINVAL === wvistau64 (64 bit string) === string.c:178: Test failed: errno = 0, expected EINVAL string.c:185: Test failed: errno = 0, expected EINVAL string.c:192: Test failed: errno = 0, expected EINVAL string.c:221: Test failed: errno = 0, expected EINVAL string.c:228: Test failed: errno = 0, expected EINVAL string.c:235: Test failed: errno = 0, expected EINVAL === w1064 (64 bit string) === string.c:178: Test failed: errno = 0, expected EINVAL string.c:185: Test failed: errno = 0, expected EINVAL string.c:192: Test failed: errno = 0, expected EINVAL string.c:221: Test failed: errno = 0, expected EINVAL string.c:228: Test failed: errno = 0, expected EINVAL string.c:235: Test failed: errno = 0, expected EINVAL

1 0

Re: [PATCH] msvcrt: Implement _memicmp_l()
by Marvin Oct. 23, 2017

Oct. 23, 2017

Hi, While running your changed tests on Windows, 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=33712 Your paranoid android. === wxppro (32 bit string) === 09f0:string: unhandled exception c0000005 at 77C4835C === w2003std (32 bit string) === 05f0:string: unhandled exception c0000005 at 77BD929C

1 0

Re: [PATCH] comctl32: Add RegisterClassNameW stub
by Nikolay Sivov Oct. 22, 2017

Oct. 22, 2017

I don't think you need separate export to register classes. As a test create your own dll with such export, and see if it's called from test exe when some referenced class name is requested, as far as I remember this was only used in early versions.

2 3

Re: [PATCH 2/2] msi: Don't leak memory on failure.
by Nikolay Sivov Oct. 22, 2017

Oct. 22, 2017

On 21.10.2017 12:50, Pierre Schweitzer wrote: > row = MSI_CreateRecord(1); > if (!row) > + { > + msi_free(buffer); > return NULL; > + } > MSI_RecordSetStringW(row, 1, buffer); > msi_free(buffer); > return row; You can replace this with: --- row = MSI_CreateRecord(1); if (row) MSI_RecordSetStringW(row, 1, buffer); msi_free(buffer); return row; ---

2 2

Re: [PATCH 5/5] msi: Add more feature action tests
by Marvin Oct. 20, 2017

Oct. 20, 2017

Hi, While running your changed tests on Windows, 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=33689 Your paranoid android. === w2003std (32 bit package) === package.c:3763: Test failed: Expected ERROR_SUCCESS, got 1603 === w7u (32 bit package) === The task timed out

1 0

Re: [PATCH 4/5] msi: Add install test with no arguments on already installed product
by Marvin Oct. 20, 2017

Oct. 20, 2017

Hi, While running your changed tests on Windows, 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=33688 Your paranoid android. === w2003std (32 bit package) === package.c:3710: Test failed: Expected ERROR_SUCCESS, got 1603

1 0

Re: [resend PATCH 4/4] gdiplus/metafile: Implement EmfPlusRecordTypeSetClipRegion playback
by Vincent Povirk Oct. 20, 2017

Oct. 20, 2017

> + return GdipSetClipRegion(real_metafile->playback_graphics, real_metafile->objtable[region].u.region, > + (flags >> 8) & 0xf); This clip region needs to be intersected with the clip region from when the metafile began playing, similar to what we do for world transforms. Alternately, we could try to repurpose our existing code for handling gdi32 clipping and transforms. It just occurred to me that applications could tell the difference between those approaches, so we should probably have tests.

1 0

Re: [PATCH 12/14] mshtml: Don't use fire_event to dispatch script readystatechange event.
by Marvin Oct. 20, 2017

Oct. 20, 2017

Hi, While running your changed tests on Windows, 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=33685 Your paranoid android. === w7u (32 bit htmldoc) === htmldoc.c:2682: Test failed: unexpected call GetOverrideKeyPath

1 0

Re: [01/10] secur32: Implement AcquireCredentialsHandle for Kerberos.
by Dmitry Timoshkov Oct. 20, 2017

Oct. 20, 2017

Hi Hans, Hans Leidekker <hans(a)codeweavers.com> wrote: > This series is based on a patch by Rob Shearman. Since during some time I was involved in the project that aims adding Kerberos support to Wine (it's still not fully finished yet, that's why we haven't published it yet) I have some comments on your approach. Our main target for testing was KerberosAuthenticationTester.exe from http://blog.michelbarneveld.nl/michel/archive/2009/12/05/kerberos-authentic… (the author doesn't provide the sources, but it's pretty trivial to get full C# source for the whole application, so understanding and debugging is not an issue). Another good test is klist.exe from (older) PSDK. Our team also decided to use Rob's patch as a base for SSP, and in addition we've also implemented Kerberos ticket management API. Main difference between your and our approaches is the architecture. We decided to add support for Kerberos using an approximation to what Windows has in place: implement Authentication Package (AP) manager in secur32 (LSA APIs) which dinamically loads APs listed in the registry. All the real code (both SSP and AP) is implemented in kerberos.dll, which dynamically loads libkrb5.so and gssapi_krb5.so. Did you consider moving the Kerberos support into separate kerberos AP/SSP dll? -- Dmitry.

2 16

Re: [2/3] ole32/tests: Add a test for IOleCache storage contents created by IPersistStorage_Save(). (v2)
by Huw Davies Oct. 20, 2017

Oct. 20, 2017

On Fri, Oct 20, 2017 at 09:58:57AM +0800, Dmitry Timoshkov wrote: > v2: Call IPersistStorage_Save() directly instead of OleSave(). > > Signed-off-by: Dmitry Timoshkov <dmitry(a)baikal.ru> > --- > dlls/ole32/tests/ole2.c | 402 ++++++++++++++++++++++++++++++++++++++++++++++++ > 1 file changed, 402 insertions(+) > > diff --git a/dlls/ole32/tests/ole2.c b/dlls/ole32/tests/ole2.c > index 0773bbf189..9ed19f1f9b 100644 > --- a/dlls/ole32/tests/ole2.c > +++ b/dlls/ole32/tests/ole2.c > +static void check_storage_contents(IStorage *stg, const struct storage_def *stg_def, > + int *enumerated_streams, int *matched_streams) > +{ > + HRESULT hr; > + IEnumSTATSTG *enumstg; > + IStream *stream; > + STATSTG stat; > + int i, seen_stream[MAX_STREAM] = { 0 }; > + > + if (winetest_debug > 1) > + trace("check_storage_contents:\n=============================================\n"); > + > + *enumerated_streams = 0; > + *matched_streams = 0; > + > + hr = IStorage_Stat(stg, &stat, STATFLAG_NONAME); > + ok(hr == S_OK, "unexpected %#x\n", hr); > +todo_wine_if(!IsEqualCLSID(stg_def->clsid, &stat.clsid)) > + ok(IsEqualCLSID(stg_def->clsid, &stat.clsid), "expected %s, got %s\n", > + wine_dbgstr_guid(stg_def->clsid), wine_dbgstr_guid(&stat.clsid)); > + > + hr = IStorage_EnumElements(stg, 0, NULL, 0, &enumstg); > + ok(hr == S_OK, "unexpected %#x\n", hr); > + > + for (;;) > + { > + ULONG bytes; > + int clipformat = -1; > + PresentationDataHeader header; > + char name[32]; > + BYTE data[256]; > + > + memset(&header, 0, sizeof(header)); > + > + hr = IEnumSTATSTG_Next(enumstg, 1, &stat, NULL); > + if(hr == S_FALSE) break; > + ok(hr == S_OK, "unexpected %#x\n", hr); > + > + if (winetest_debug > 1) > + trace("name %s, type %u, size %d, clsid %s\n", > + wine_dbgstr_w(stat.pwcsName), stat.type, stat.cbSize.u.LowPart, wine_dbgstr_guid(&stat.clsid)); > + > + ok(stat.type == STGTY_STREAM, "unexpected %#x\n", stat.type); > + > + WideCharToMultiByte(CP_ACP, 0, stat.pwcsName, -1, name, sizeof(name), NULL, NULL); > + > + hr = IStorage_OpenStream(stg, stat.pwcsName, NULL, STGM_READ | STGM_SHARE_EXCLUSIVE, 0, &stream); > + ok(hr == S_OK, "unexpected %#x\n", hr); > + > + if (!memcmp(name, "\2OlePres", 7)) > + { > + clipformat = read_clipformat(stream); > + > + hr = IStream_Read(stream, &header, sizeof(header), &bytes); > + ok(hr == S_OK, "unexpected %#x\n", hr); > + ok(bytes >= 24, "read %u bytes, expected to read %u bytes\n", bytes, sizeof(header)); > + You're using sizeof as an argument to a fprintf-like function which, for example, gives a warning on macOS. ole2.c:3532:82: warning: format specifies type 'unsigned int' but the argument has type 'unsigned long' [-Wformat] Also, while you're at it, couldn't that '24' be sizeof(header) instead? Huw.

2 2

Re: [3/3] ole32: Remove an optimization from DataCache_Save() that copies whole original storage.
by Marvin Oct. 20, 2017

Oct. 20, 2017

Hi, While running your changed tests on Windows, 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=33675 Your paranoid android. === wxppro (32 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w2003std (32 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === wvistau64 (32 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w2008s64 (32 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w7u (32 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w7pro64 (32 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w8 (32 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w864 (32 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w1064 (32 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === wvistau64 (64 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w2008s64 (64 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w7pro64 (64 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w864 (64 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes === w1064 (64 bit ole2) === ole2.c:3529: Test failed: read 24 bytes ole2.c:3529: Test failed: read 24 bytes

1 0

Re: [2/3] ole32/tests: Add a test for IOleCache storage contents created by IPersistStorage_Save(). (v3)
by Marvin Oct. 20, 2017

Oct. 20, 2017

Hi, While running your changed tests on Windows, 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=33674 Your paranoid android. === wxppro (32 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w2003std (32 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === wvistau64 (32 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w2008s64 (32 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w7u (32 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w7pro64 (32 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w8 (32 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w864 (32 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w1064 (32 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === wvistau64 (64 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w2008s64 (64 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w7pro64 (64 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w864 (64 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes === w1064 (64 bit ole2) === ole2.c:3532: Test failed: read 24 bytes ole2.c:3532: Test failed: read 24 bytes

1 0

Re: [PATCH 1/4] gdiplus/metafile: Partially implement EmfPlusPen support for playback
by Vincent Povirk Oct. 19, 2017

Oct. 19, 2017

Some of the size calculations don't seem to match. + if (pendata->PenDataFlags & PenDataTransform) + offset += sizeof(GpMatrix); + if (pendata->PenDataFlags & PenDataTransform) + { + FIXME("PenDataTransform is not supported.\n"); + offset += sizeof(EmfPlusTransformMatrix); + } + if (pendata->PenDataFlags & PenDataDashedLine) + { + EmfPlusDashedLineData *dashedline = (EmfPlusDashedLineData *)((BYTE *)data + offset); + + offset += FIELD_OFFSET(EmfPlusDashedLineData, data); + if (data_size <= offset) + return InvalidParameter; + + offset += dashedline->DashedLineDataSize; + } + if (pendata->PenDataFlags & PenDataDashedLine) + { + EmfPlusDashedLineData *dashedline = (EmfPlusDashedLineData *)((BYTE *)pendata + offset); + FIXME("PenDataDashedLine is not supported.\n"); + offset += dashedline->DashedLineDataSize; + }

2 1

Font rendering issues after some update.
by John Found Oct. 19, 2017

Oct. 19, 2017

After an update of the system, WINE stopped to render the fonts properly. Regardless of the registry or other system settings, following behaviour is observed: 1. The antialiasing is only grayscale. 2. The antialiasing is only if the font size is bigger than 14px. Smaller fonts are not smoothed at all. 3. The hinting seems to be switched off. At least for some fonts. The bitmap fonts or the fonts designed to not use smoothing (for example FixedSys Excelsior) look like usual. All this seems to not depend on any settings of the system configuration or WINE registry. Thinking that the problem is simply broken WINE/fontconfig/freetype I tried all possible configuration tweaks from Internet. Without any effect. I mean, the font rendering never changes even by pixel. I have second system, that worked fine, but after the today update (I am using it very rarely) the problem appeared as well. In the same time, all other Linux applications work normally, I can change the rendering options from the usual GUI applets. I can't say when exactly the problem appeared for the first time. But on this second system, the updated versions are: wine: from 2.4-1 to 2.18-1 fontconfig: from 2.12.1-4 to 2.12.6-1 freetype: from 2.7.1-1 to 2.8.1-1 Any ideas how to fix the problem? Or at least how to continue the research. -- http://fresh.flatassembler.net http://asm32.info John Found <johnfound(a)asm32.info>

2 3

Re: [PATCH 3/9] mshtml: Set current event target in fire_event_obj.
by Marvin Oct. 19, 2017

Oct. 19, 2017

Hi, While running your changed tests on Windows, 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=33648 Your paranoid android. === w7u (32 bit htmldoc) === htmldoc.c:2682: Test failed: unexpected call GetOverrideKeyPath

1 0

Re: [PATCH 2/2] user32: Implement IDI_SHIELD.
by Alexandre Julliard Oct. 19, 2017

Oct. 19, 2017

Alex Henrie <alexhenrie24(a)gmail.com> writes: > From 28d5fcc630dcf3ee9b559f05a9a0fb5d6c3e06bd Mon Sep 17 00:00:00 2001 > From: Alex Henrie <alexhenrie24(a)gmail.com> > Date: Thu, 19 Oct 2017 01:30:55 -0600 > Subject: [PATCH 2/2] user32: Implement IDI_SHIELD. > > Fixes https://bugs.winehq.org/show_bug.cgi?id=43770 > > Steps to create these icons: Thanks! While you are at it, how about adding some colors to it? On a gray background it's going to be pretty much invisible. -- Alexandre Julliard julliard(a)winehq.org

2 1

Re: [PATCH 1/4] d3d9/tests: Windows 10 17.09 added an undocumented query type (v2).
by Henri Verbeet Oct. 19, 2017

Oct. 19, 2017

On 19 October 2017 at 10:09, Stefan Dösinger <stefan(a)codeweavers.com> wrote: > @@ -5552,7 +5552,9 @@ static void test_query_support(void) > > for (i = 0; i < 40; ++i) > { > - if (D3DQUERYTYPE_VCACHE <= i && i <= D3DQUERYTYPE_MEMORYPRESSURE) > + /* Windows 10 17.09 (build 16299.19) added an undocumented query with an enum value of 0x16 (=22). > + * It returns D3D_OK when asking for support and E_FAIL when trying to actually create it. */ > + if (D3DQUERYTYPE_VCACHE <= i && i <= D3DQUERYTYPE_MEMORYPRESSURE || i == 0x16) This introduces a warning: device.c: In function ‘test_query_support’: device.c:5566:38: warning: suggest parentheses around ‘&&’ within ‘||’ [-Wparentheses] if (D3DQUERYTYPE_VCACHE <= i && i <= D3DQUERYTYPE_MEMORYPRESSURE || i == 0x16)

1 0

Re: [PATCH 2/2] msi: Remove full reinstall special handling
by Hans Leidekker Oct. 19, 2017

Oct. 19, 2017

On Wed, 2017-10-18 at 17:25 +0200, Piotr Caban wrote: > This patch fixes "change office 2010 serial key" functionality. This code was introduced to fix the Enterprise Architect installer after I broke it with e2900c23c7f0b143078da780c9a0ff19a895bf45. (see https://bugs.winehq.org/show_bug.cgi?id=32387) Your patch breaks this installer again because the condition for entering the maintenance dialog tests the Preselected property. Given your test it seems that we shouldn't set REINSTALL at all (since it causes Preselected to be set) or that we should set it only after evaluating command line overrides.

2 1

Re: [3/3] ole32: Use a shortcut in DataCache_Save only if fSameAsLoad set to TRUE.
by Huw Davies Oct. 19, 2017

Oct. 19, 2017

On Wed, Sep 20, 2017 at 02:54:21PM +0300, Dmitry Timoshkov wrote: > This looks like a typo, and this patch makes saving data in a pretty complex > application work. > > Signed-off-by: Dmitry Timoshkov <dmitry(a)baikal.ru> > --- > dlls/ole32/datacache.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/dlls/ole32/datacache.c b/dlls/ole32/datacache.c > index 4e684b4189..c3383176f8 100644 > --- a/dlls/ole32/datacache.c > +++ b/dlls/ole32/datacache.c > @@ -1663,7 +1663,7 @@ static HRESULT WINAPI DataCache_Save( > } > > /* this is a shortcut if nothing changed */ > - if (!dirty && !fSameAsLoad && This->presentationStorage) > + if (!dirty && fSameAsLoad && This->presentationStorage) > { > return IStorage_CopyTo(This->presentationStorage, 0, NULL, NULL, pStg); > } No, this isn't a typo. The idea is to copy the loaded storage to a fresh storage that is passed when fSameAsLoad is FALSE. I'm aware that there are issues related to the storage handling in both the cache and the default handler. That's on my list of things to look into, but feel free to investigate further. Huw.

2 7

Re: [1/2] ole32/tests: Add a test for OleSave() called for a IOleCache storage.
by Huw Davies Oct. 19, 2017

Oct. 19, 2017

On Thu, Oct 19, 2017 at 01:06:50PM +0800, Dmitry Timoshkov wrote: > Signed-off-by: Dmitry Timoshkov <dmitry(a)baikal.ru> > --- > dlls/ole32/tests/ole2.c | 214 +++++++++++++++++++++++++++++++++++++++++++++--- > 1 file changed, 202 insertions(+), 12 deletions(-) This looks ok, but could you please call IPersistStorage_Save() directly rather than using the OleSave() helper? Also fix the function name to reflect that. Thanks, Huw.

2 2

Re: [2/2] ole32/tests: Add a test for IOleCache storage contents created by OleSave().
by Huw Davies Oct. 19, 2017

Oct. 19, 2017

On Thu, Oct 19, 2017 at 01:07:01PM +0800, Dmitry Timoshkov wrote: > Signed-off-by: Dmitry Timoshkov <dmitry(a)baikal.ru> > --- > dlls/ole32/tests/ole2.c | 372 ++++++++++++++++++++++++++++++++++++++++++++++++ > 1 file changed, 372 insertions(+) > Again, please use IPersistStorage_Save() in this one. Huw.

1 0