[CRIU] [PATCH]v2 zdtm: test semi-closed unix stream connection for data loss

[Ruslan Kuprieiev]

On 08/27/2013 12:34 PM, Andrew Vagin wrote:
> On Tue, Aug 27, 2013 at 12:21:09PM +0400, Ruslan Kuprieiev wrote:
>> On 08/27/2013 08:28 AM, Andrew Vagin wrote:
>>> On Tue, Aug 27, 2013 at 01:38:42AM +0400, Ruslan Kuprieiev wrote:
>>>> On 08/27/2013 01:07 AM, Andrew Vagin wrote:
>>>>> On Mon, Aug 26, 2013 at 09:27:27PM +0400, Ruslan Kuprieiev wrote:
>>>>>> Hi!
>>>>>>
>>>>>> Added test for data loss in semi-closed stream unix connection.
>>>>> Why do you decide to add a new test? In a previous attempt you tryed to
>>>>> expand the socket02 test and I think it was the right way.
>>>>> You replaced one check on another and I suggested to add one more check
>>>>> and do nothing with the previous one.
>>>>>
>>>> It was impulsive decision=).
>>>> Here is one more try.
>>>> If this patch is ok, please, tell me if i should send it again with
>>>> [PATCH]v3 tag.
>>>> Signed-off-by: Ruslan Kuprieiev <kupruser at gmail.com>
>>>>
>>>> diff --git a/test/zdtm/live/static/sockets02.c b/test/zdtm/live/static/sockets02.c
>>>> index 8ff38f8..3e6829d 100644
>>>> --- a/test/zdtm/live/static/sockets02.c
>>>> +++ b/test/zdtm/live/static/sockets02.c
>>>> @@ -12,6 +12,7 @@
>>>>   #include <sys/stat.h>
>>>>   #include <limits.h>
>>>>   #include <fcntl.h>
>>>> +#include <time.h>
>>>>   #include "zdtmtst.h"
>>>> @@ -21,7 +22,10 @@ const char *test_author	= "Pavel Emelyanov <xemul at parallels.com>\n";
>>>>   int main(int argc, char *argv[])
>>>>   {
>>>>   	int ssk_pair[2], ret;
>>>> -	char aux;
>>>> +	char aux, data;
>>>> +
>>>> +	srandom(time(NULL));
>>>> +	data = (char)random();
>>>>   	test_init(argc, argv);
>>>> @@ -30,6 +34,12 @@ int main(int argc, char *argv[])
>>>>   		exit(1);
>>>>   	}
>>>> +	ret = write(ssk_pair[1], &data, sizeof(data));
>>>> +	if (ret != sizeof(data)) {
>>>> +		fail("write\n");
>>>> +		exit(1);
>>>> +	}
>>>> +
>>>>   	close(ssk_pair[1]);
>>>>   	test_daemon();
>>>> @@ -37,9 +47,12 @@ int main(int argc, char *argv[])
>>> We need to check data and the state of socket. I think we can add a
>>> separate check for data.
>> I'm quite confused=). Do you want me to create new socket pair and
>> test data loss on it, or you want me to add receiving data and
>> checking it before if(ret != 0 || errno !=0) ?
> I suggest to check data before checking that socket is closed. I wrote
> pseudo-patch in the previous message.
>
> 	... check data ...
>
>          errno = 0;
>          ret = read(ssk_pair[0], &aux, sizeof(aux));
Oh, sorry, i didn't notice. Thanks.