Re: [PATCH 1/2] dinput/tests: Do not test that reports are identical when polling.

[Rémi Bernon]

On 5/2/22 23:48, Zebediah Figura wrote:
> This is dependent on timing, and currently fails occasionally both on Windows
> and Wine.
> Signed-off-by: Zebediah Figura
>
> For examples of failures:
>
> http://test.winehq.org/data/4ec67b7a6447dfc4af8c03c141c600b41b90ef53/win81_n...
> http://test.winehq.org/data/64b96eec7d0aea470f897a3ed0ac9e1b3a680cc5/linux_f...
>
> dlls/dinput/tests/hid.c | 6 ++----
>   1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/dlls/dinput/tests/hid.c b/dlls/dinput/tests/hid.c
> index 4e107625064..1162f154f8e 100644
> --- a/dlls/dinput/tests/hid.c
> +++ b/dlls/dinput/tests/hid.c
> @@ -2312,14 +2312,13 @@ static void test_hidp( HANDLE file, HANDLE async_file, int report_id, BOOL polle
>           ok( ret, "GetOverlappedResult failed, last error %lu\n", GetLastError() );
>           ok( value == (report_id ? 3 : 4), "GetOverlappedResult returned length %lu, expected %u\n",
>               value, (report_id ? 3 : 4) );
> -
>    /* first report should be ready and the same */
> -
>    /* first report should be ready */
>     ret = GetOverlappedResult( async_file, &overlapped, &value, FALSE );
>     ok( ret, "GetOverlappedResult failed, last error %lu\n", GetLastError() );
>     ok( value == (report_id ? 3 : 4), "GetOverlappedResult returned length %lu, expected %u\n",
>         value, (report_id ? 3 : 4) );
>     ok( memcmp( report, buffer + caps.InputReportByteLength, caps.InputReportByteLength ),
>         "expected different report\n" );
> -
>    ok( !memcmp( report, buffer, caps.InputReportByteLength ), "expected identical reports\n" );
>
>     value = 10;
>     SetLastError( 0xdeadbeef );

This defeats a bit the purpose of the test, which is to validate that in 
polled mode all pending reads should be completed at once when device 
poll completes, whereas in non-polled mode, only one pending read should 
complete for each received report.

I don't think this one is failing so often.
> @@ -2343,12 +2342,11 @@ static void test_hidp( HANDLE file, HANDLE async_file, int report_id, BOOL polle
>           ok( ret, "GetOverlappedResult failed, last error %lu\n", GetLastError() );
>           ok( value == (report_id ? 3 : 4), "GetOverlappedResult returned length %lu, expected %u\n",
>               value, (report_id ? 3 : 4) );
> -
>    /* first report should be ready and the same */
> -
>    /* first report should be ready */
>     ret = GetOverlappedResult( async_file, &overlapped, &value, FALSE );
>     ok( ret, "GetOverlappedResult failed, last error %lu\n", GetLastError() );
>     ok( value == (report_id ? 3 : 4), "GetOverlappedResult returned length %lu, expected %u\n",
>         value, (report_id ? 3 : 4) );
> -
>    ok( !memcmp( report, buffer, caps.InputReportByteLength ), "expected identical reports\n" );
>
>     send_hid_input( file, expect_small, sizeof(expect_small) );

I'm not completely sure what this second one is trying to achieve, but 
probably comparing the reports there makes indeed little sense, as the 
polling frequency is increased a lot, and they will likely often differ.