Re: [LTP] [PATCH v2 2/3] lib/tst_res.c: introduce function tst_record_childstatus()

[<ch...@su...>]

Hi!
> diff --git a/lib/tst_res.c b/lib/tst_res.c
> index 079fc42..24fb4d9 100644
> --- a/lib/tst_res.c
> +++ b/lib/tst_res.c
> @@ -101,6 +101,9 @@
>  #include <stdarg.h>
>  #include <string.h>
>  #include <unistd.h>
> +#include <sys/types.h>
> +#include <sys/wait.h>
> +
>  #include "test.h"
>  #include "usctest.h"
>  #include "ltp_priv.h"
> @@ -228,6 +231,40 @@ void tst_reset_exitval(void)
>  {
>  	T_exitval = 0;
>  }
> +
> +int tst_record_childstatus(int status, int *sig)
> +{
> +	int ret = 0, ttype_result, s = 0;
> +
> +	if (WIFEXITED(status)) {
> +		ret = WEXITSTATUS(status);
> +		ttype_result = TTYPE_RESULT(ret);
> +		T_exitval |= ttype_result;
> +		s = 0;
> +		ret = CHILD_EXIT_NORMALLY;
> +	} else if (WIFSIGNALED(status)) {
> +		s = WTERMSIG(status);
> +		tst_resm(TINFO, "child process is killed by signal: %s(%d)",
> +			 tst_strsig(s), s);
> +		ret = CHILD_EXIT_SIGNALED;

I would not bother with the return value and just do tst_brkm() with
TBROK in the rest of the cases. Because child killed by a signal is a
very special case and testcases that needs to assert it can do the
wait() and WIFSIGNALED() themselves.

-- 
Cyril Hrubis
ch...@su...