[LTP] [PATCH 2/3] syscalls/getrusage: Cleanup and bugfix for getrusage03

[Cyril Hrubis]

Hi!
> Same with lib: tst_process_state: Add tst_process_release_wait().
> 
> If we call the signal(SIGCHLD,SIG_IGN), the SIGCHLD signal is ignored by 
> the system. Thus, no zombie is created before the child is terminated. 
> The logs are as follows:
> 
> getrusage03.h:27: TINFO: allocate 400MB
> getrusage03.c:39: TPASS: check that initial.children ~= pre_wait.children
> getrusage03.c:39: TPASS: check that post_wait.children ~= 400MB
> getrusage03.h:27: TINFO: allocate 500MB
> getrusage03.c:123: TBROK: Failed to open FILE '/proc/84598/stat' for 
> reading: ENOENT (2)
> 
> So I write TST_PROCESS_RELEASE_WAIT() here to check /proc/$PID.

Ah right, reading the documentation if parent ignores sigchild the child
is discarded and there is no trace of it on the system once it returns
from main().

So I guess that we should add a function that checks that a given pid is
present on a system but there is simpler interface for this, if you do
kill(2) with signal == 0 it returns with with errno set to ESRCH if the
pid is no longer present on a system.

Also I'm not sure if release is right word here, maybe it would better
describe the action if the function name was tst_process_exit_wait().

> I noticed that some safe_ micro checks tst_test->forks_child or other 
> tst_test->xx, while tst_test is not defined in some functions with 
> #define TST_NO_DEFAULT_MAIN.

Ah right, I guess that in a case of fork() and clone() calls it would be
easier not to use SAFE_MACROS() for processes that have been started by
exec().

-- 
Cyril Hrubis
chrubis@suse.cz