[PATCH 08/11] kernelbase/console, programs/conhost: support CONSOLE_READCONSOLE_CONTROL in ReadConsoleW

[Eric Pouech]

Le 20/02/2022 à 14:03, Jacek Caban a écrit :
> Hi Eric,
>
> On 2/9/22 16:44, Eric Pouech wrote:
>> Signed-off-by: Eric Pouech<eric.pouech at gmail.com>
>>
>> ---
>>   dlls/kernelbase/console.c  |   32 ++++++++++++++-
>>   include/wine/condrv.h      |    1
>>   programs/conhost/conhost.c |   93 
>> +++++++++++++++++++++++++++++++++++---------
>>   programs/conhost/conhost.h |    1
>>   server/console.c           |    1
>>   5 files changed, 107 insertions(+), 21 deletions(-)
>>
>> diff --git a/dlls/kernelbase/console.c b/dlls/kernelbase/console.c
>> index a7eeb439232..52949ab7541 100644
>> --- a/dlls/kernelbase/console.c
>> +++ b/dlls/kernelbase/console.c
>> @@ -1598,8 +1598,36 @@ BOOL WINAPI ReadConsoleW( HANDLE handle, void 
>> *buffer, DWORD length, DWORD *coun
>>           return FALSE;
>>       }
>>   -    ret = console_ioctl( handle, IOCTL_CONDRV_READ_CONSOLE, NULL, 
>> 0, buffer,
>> -                         length * sizeof(WCHAR), count );
>> +    if (reserved)
>> +    {
>> +        CONSOLE_READCONSOLE_CONTROL* crc = reserved;
>> +        char *tmp;
>> +
>> +        if (crc->nLength != sizeof(*crc) || crc->nInitialChars >= 
>> length)
>> +        {
>> +            SetLastError( ERROR_INVALID_PARAMETER );
>> +            return FALSE;
>> +        }
>> +        if (!(tmp = HeapAlloc( GetProcessHeap(), 0, sizeof(DWORD) + 
>> crc->nInitialChars * sizeof(WCHAR) )))
>> +        {
>> +            SetLastError( ERROR_NOT_ENOUGH_MEMORY );
>> +            return FALSE;
>> +        }
>> +
>> +        memcpy( tmp, &crc->dwCtrlWakeupMask, sizeof(DWORD) );
>> +        memcpy( tmp + sizeof(DWORD), buffer, crc->nInitialChars * 
>> sizeof(WCHAR) );
>> +        ret = console_ioctl( handle, IOCTL_CONDRV_READ_COMPLETION,
>> +                             tmp, sizeof(DWORD) + crc->nInitialChars 
>> * sizeof(WCHAR),
>> +                             buffer, length * sizeof(WCHAR),
>> +                             count );
>> +        crc->dwConsoleKeyState = 0;
>> +        HeapFree( GetProcessHeap(), 0, tmp );
>> +    }
>> +    else
>> +    {
>> +        ret = console_ioctl( handle, IOCTL_CONDRV_READ_CONSOLE, 
>> NULL, 0, buffer,
>> +                             length * sizeof(WCHAR), count );
>> +    }
>>       if (ret) *count /= sizeof(WCHAR);
>>       return ret;
>
>
> Could we just use IOCTL_CONDRV_READ_CONSOLE in both cases and 
> distinguish the actual mode by checking input size in conhost?

that's possible ;-) (first version of the patch were written that way)

but there's a different structure on input and on output between the two 
modes. We need to differentiate anyway in conhost between the two modes.

I thought it was clearer this way

I also considered the other way around (only using the new mode) but 
thought the extra copy on return wasn't worth it

>
>
> Also, it would be interesting to try to add a test for this, see 
> conhost/tests/tty.c for an example of other ReadConsole() tests. A 
> fair warning: it may prove to be tricky. When I wrote those tests, I 
> observed some weird or inconsistent Windows behaviour. Pseudo consoles 
> were not matured on Windows back then, hopefully things will be better 
> now. There is a huge chance that it's not one of those cases and it 
> will be straightforward. If things get ugly, I'm fine skipping tests, 
> but it's worth to try.
ok I'll give it a try
>
>