[LTP] [v5,2/2] semop04: Refactor with new API

Richard Palethorpe rpalethorpe@suse.de
Tue Aug 29 11:20:44 CEST 2023


Hello,

Petr Vorel <pvorel@suse.cz> writes:

> Hi Wei,
>
>> -void delayloop()
>> +static void delayloop(void)
>>  {
>>  	int delay;
>> +
>>  	delay = 1 + ((100.0 * rand()) / RAND_MAX);
>> -	if (verbose)
>> -		printf("in delay function for %d microseconds\n", delay);
>>  	usleep(delay);
>>  }
>
> ...
>> +	for (i = 0; i < LOOPS; i++) {
>> +		semdown(semid);
>>  		delayloop();
> Why not just:
>
> 	usleep(1 + ((100.0 * rand()) / RAND_MAX));
>
> and delete whole delayloop() function?
>
>> -		if (semup(semid)) {
>> -			printf("semup failed\n");
>> -		}
>> -		if (verbose)
>> -			printf("sem is up\n");
>> +		semup(semid);
>>  	}
>
> ...
>>  	/* set up the semaphore */
>> -	if ((semid = semget((key_t) 9142, 1, 0666 | IPC_CREAT)) < 0) {
>> -		printf("error in semget()\n");
>> -		exit(-1);
>> -	}
>> +	semid = SAFE_SEMGET((key_t) 9142, 1, 0666 | IPC_CREAT);
>> +
>>  	semunion.val = 1;
>> -	if (semctl(semid, 0, SETVAL, semunion) == -1) {
>> -		printf("error in semctl\n");
>> -	}
>
>> -	if ((pid = fork()) < 0) {
>> -		printf("fork error\n");
>> -		exit(-1);
>> -	}
>> +	SAFE_SEMCTL(semid, 0, SETVAL, semunion);
>> +
>> +	pid = SAFE_FORK();
>> +
>>  	if (pid) {
>> -		/* parent */
>>  		srand(pid);
>>  		mainloop(semid);
>> -		waitpid(pid, &chstat, 0);
>> -		if (!WIFEXITED(chstat)) {
>> -			printf("child exited with status\n");
>> -			exit(-1);
>> -		}
>> -		if (semctl(semid, 0, IPC_RMID, semunion) == -1) {
>> -			printf("error in semctl\n");
>> -		}
>> -		if (errors) {
>> -			printf("FAIL: there were %d errors\n", errors);
>> -		} else {
>> -			printf("PASS: error count is 0\n");
>> -		}
>> -		exit(errors);
>> +		tst_reap_children();
>> +		SAFE_SEMCTL(semid, 0, IPC_RMID, semunion);
>> +		tst_res(TPASS, "Semaphore up/down check success");
>
> Maybe here use plain semctl() verified with TST_EXP_POSITIVE() from
> tst_test_macros.h? This way we'd get TFAIL on failure instead of TPASS.
>
> Kind regards,
> Petr

Setting to changes requested in patchwork.


-- 
Thank you,
Richard.


More information about the ltp mailing list