[Perfparse-devel-int] Re: [Perfparse-users] perfparsed & CPU usage

[Yves <yme...@pe...>]

> Yves,
>
> I've done an strace, results are attached. Changing &tm to NULL did
> not change anything.

Well, here is what your perfparsed does when doing nothing :

select(6, [4 5], NULL, NULL, NULL)      =3D 1 (in [4])
rt_sigprocmask(SIG_BLOCK, [INT TERM], [], 8) =3D 0
read(4, "", 10)                         =3D 0
rt_sigprocmask(SIG_SETMASK, [], NULL, 8) =3D 0
wait4(-1, NULL, WNOHANG, NULL)          =3D -1 ECHILD (No child processes=
)

In other words, there is always something to read on the 4th file descrip=
tor (1st line),
even if we see on the 3rd line that it reads nothing.

There is an additionnal test that you could maybe do, still in log_reader=
.c. You can
find in open_log_source a big test :
if(!strcmp(filename,"-")) {
   ...
} else if(filename[0] =3D=3D '>') {
   ...
} else if(filename[0] =3D=3D '|') {
   ...
   log_source_nonblock(i);
   ...
} else {
   ...
}


Could you comment that log_source_nonblock(i), and only that one ?
I hope this is the fix. Otherwise, I will have to find why, when there is=
 nothing to
read on the pipe, why it considers there is something to read anyways.

Note : per...@li... is the mailing-list for the core =
developers,
including Ben. If you don't mind, answer with that list in CC :)

Yves


--=20
- Homepage    - http://ymettier.free.fr - http://www.logicacmg.com -
- GPG key     - http://ymettier.free.fr/gpg.txt                    -
- Maitretarot - http://www.nongnu.org/maitretarot/                 -
- Perfparse   - http://perfparse.sf.net/                           -