On Sun, Jun 15, 2008 at 10:33 AM, John Kacur <jkacur@xxxxxxxxx> wrote:
> On Sat, Jun 14, 2008 at 1:40 AM, Leon Woestenberg
> <leon.woestenberg@xxxxxxxxx> wrote:
>>
>> Hello,
>>
>> I noticed many of my tests were bogus, as somehow my "-t NUM" argument
>> is not parsed properly on my target:
>>
>> Linux efika 2.6.25.4-rt6 #10 PREEMPT RT Thu Jun 12 00:34:11 CEST 2008
>> ppc unknown
>>
>> cyclictest 0.21 (with some extra argument printing added)
>>
>> root@efika:~# /cyclictest -t 2 -n -p 50
>
> -t takes an optional argument, that means it will only be parsed
> correctly if you write it together without a space with the t, that
> is, -t2
>
>> threads = 1
>> started with interval 1000
>>
>> T: 0 ( 1178) P:50 I:1000 C: 333 Min: 45 Act: 46 Avg: 49 Max: 100
>>
>> root@efika:~# /cyclictest -t2 -n -p 50
>
> Here you wrote it together and it was parsed properly.
>
>> threads = 2
>> started with interval 1000
>> started with interval 1500
>>
>> on my x86/32 system the same happens (with -t 4 it defaults to 2
>> because it's a dual-core, with -t4 it works).
>>
>> leon@witty:/tmp$ sudo ./cyclictest -t 4 -p 50
>
> Once again, here you wrote it with a space in-between so it is not
> honoured. -p 50 on the other hand is parsed correctly because it takes
> a non-optional argument.
>> threads = 2
>
> ---SNIP---
>>
>> At this moment I am puzzled, but I suspect the optional_argument handling.
>>
> I think you are right, it is just a puzzling thing about getopt since
> it parses non-optional arguments with or without a space, but optional
> arguments must be written without a space. Quoting from the manpage.
>
> Two colons mean an option takes an optional arg; if there is text in
> the current argv-element (i.e., in the same word as the option name
> itself, for example, "-oarg"), then it is returned in optarg, otherwise
> optarg is set to zero.
>
> However, if optarg is set to zero, we could still try to parse the
> next argument using the optind, so here is my attempt to do so, can
> you test it?
>
> Index: rt/rt-tests/src/cyclictest/cyclictest.c
> ===================================================================
> --- rt.orig/rt-tests/src/cyclictest/cyclictest.c 2008-06-04
> 17:30:07.000000000 +0200
> +++ rt/rt-tests/src/cyclictest/cyclictest.c 2008-06-15
> 10:28:39.000000000 +0200
> @@ -680,7 +680,8 @@
> if (optarg != NULL)
> num_threads = atoi(optarg);
> else
> - num_threads = max_cpus;
> + if (!(num_threads = atoi(argv[optind])))
> + num_threads = max_cpus;
> break;
> case 'v': verbose = 1; break;
> case '?': error = 1; break;
>
Thinking just a bit more about this - I don't want to introduce a new
bug, if you put the -t option at the end without an argument, the
above patch could sigsegv, so the following patch is more robust,
hopefully production ready with the sign-off included
Index: rt/rt-tests/src/cyclictest/cyclictest.c
===================================================================
--- rt.orig/rt-tests/src/cyclictest/cyclictest.c 2008-06-04
17:30:07.000000000 +0200
+++ rt/rt-tests/src/cyclictest/cyclictest.c 2008-06-15
11:06:34.000000000 +0200
@@ -677,11 +677,13 @@
case 'r': timermode = TIMER_RELTIME; break;
case 's': use_system = MODE_SYS_OFFSET; break;
case 't':
- if (optarg != NULL)
- num_threads = atoi(optarg);
- else
- num_threads = max_cpus;
- break;
+ if (optarg != NULL)
+ num_threads = atoi(optarg);
+ else if (argv[optind] && atoi(argv[optind]))
+ num_threads = atoi(argv[optind]);
+ else
+ num_threads = max_cpus;
+ break;
case 'v': verbose = 1; break;
case '?': error = 1; break;
}
Signed-off-by: John Kacur <jkacur@xxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
