Re: [PATCH 1/4] comctl32/tooltips: Fix an obviously wrong structure size check (try2)

[Paul Vriens]

On 10/15/2009 12:59 PM, Nikolay Sivov wrote:
> Changelog:
>      - try2: fixed usage of W calls in tests, fixed some typos
>      - fix an obviously wrong structure size check

Hi Nikolay,

I have been checking the failures for these tests. At first I just 
wanted to add some broken() statements.

Looking at test.winehq.org I can see that the test failure output is 
different sometimes. Changing the following for every call:

ti.cbSize = TTTOOLINFO[AW]_V2_SIZE - sizeof(void*);

make the output consistent (still different from comctl32 > 5.81).

What would be the best way to fix these tests? Add a define for the size 
of the struct that differs between "_WIN32_WINNT >= 0x0501" and the 
earlier versions and then have a second possible result (or broken())?

Another thing is that this test doesn't have any ok()'s, what was the 
intention here?:

457     ti.cbSize = TTTOOLINFOA_V2_SIZE;
458     ti.uId = 0x1234ABCD;
459     ti.lParam = 0xaaaaaaaa;
460     r = SendMessageA(hwnd, TTM_SETTOOLINFOA, 0, (LPARAM)&ti);

-- 
Cheers,

Paul.

[PATCH 1/4] comctl32/tooltips: Fix an obviously wrong structure size check (try2)

[Nikolay Sivov]

Paul Vriens wrote:
> On 10/15/2009 12:59 PM, Nikolay Sivov wrote:
> > Changelog:
> >      - try2: fixed usage of W calls in tests, fixed some typos
> >      - fix an obviously wrong structure size check
>
> Hi Nikolay,
>
> I have been checking the failures for these tests. At first I just 
> wanted to add some broken() statements.
>
> Looking at test.winehq.org I can see that the test failure output is 
> different sometimes. Changing the following for every call:
>
> ti.cbSize = TTTOOLINFO[AW]_V2_SIZE - sizeof(void*);
>
> make the output consistent (still different from comctl32 > 5.81).
> This makes TTTOOLINFO[AW]_V2_SIZE - sizeof(void*) == TTTOOLINFO[AW]_V1_SIZE.
> So test doesn't make sense after such modification. It's better to mark 
> it broken. This size check doesn't seem to be consistent
> at all.
>
> What would be the best way to fix these tests? Add a define for the 
> size of the struct that differs between "_WIN32_WINNT >= 0x0501" and 
> the earlier versions and then have a second possible result (or 
> broken())?
> For example for Win95 we should skip the rest of the test with V2 and 
> higher after this failure:

432     ti.cbSize = TTTOOLINFOA_V2_SIZE;
 433     ti.hwnd = NULL;
 434     ti.hinst = GetModuleHandleA(NULL);
 435     ti.uFlags = 0;
 436     ti.uId = 0x1234ABCD;
 437     ti.lpszText = NULL;
 438     ti.lParam = 0xdeadbeef;
 439     GetClientRect(hwnd, &ti.rect);
 440     r = SendMessageA(hwnd, TTM_ADDTOOLA, 0, (LPARAM)&ti);
 441     ok(r, "Adding the tool to the tooltip failed\n");
> tooltips.c:441 
> <http://source.winehq.org/git/wine.git/?a=blob;f=dlls/comctl32/tests/tooltips.c;hb=8c743cb25f2d2c6257f0786b887422d485e84eb3#l441>: 
> Test failed: Adding the tool to the tooltip failed
> it indicates that V2 isn't supported.
>
> Another thing is that this test doesn't have any ok()'s, what was the 
> intention here?:
>
> 457     ti.cbSize = TTTOOLINFOA_V2_SIZE;
> 458     ti.uId = 0x1234ABCD;
> 459     ti.lParam = 0xaaaaaaaa;
> 460     r = SendMessageA(hwnd, TTM_SETTOOLINFOA, 0, (LPARAM)&ti);
> It's just to revert back to 0xaaaaaaaa lParam, not very useful though. 
> MSDN tells it doesn't return any reasonable
> in this case, so feel free to remove assignment to r.