[LTP] [PATCH] pan/ltp-pan.c: fix incorrect number of total tests

[LTP] [PATCH] pan/ltp-pan.c: fix incorrect number of total tests

[Xiao Yang]

The number of total tests should indicate the number of tests
which have been run, but it indicates the number of tests defined
in command-file.

Signed-off-by: Xiao Yang <yangx.jy@cn.fujitsu.com>
---
 pan/ltp-pan.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/pan/ltp-pan.c b/pan/ltp-pan.c
index 1614c70..beed623 100644
--- a/pan/ltp-pan.c
+++ b/pan/ltp-pan.c
@@ -645,7 +645,7 @@ int main(int argc, char **argv)
 				strerror(errno));
 		fprintf(logfile,
 			"\n-----------------------------------------------\n");
-		fprintf(logfile, "Total Tests: %d\n", coll->cnt);
+		fprintf(logfile, "Total Tests: %d\n", (coll->cnt - starts));
 		fprintf(logfile, "Total Skipped Tests: %d\n", tconfcnt);
 		fprintf(logfile, "Total Failures: %d\n", failcnt);
 		fprintf(logfile, "Kernel Version: %s\n", unamebuf.release);
-- 
1.8.3.1

[LTP] [PATCH] pan/ltp-pan.c: fix incorrect number of total tests

[Cyril Hrubis]

Hi!
>  pan/ltp-pan.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/pan/ltp-pan.c b/pan/ltp-pan.c
> index 1614c70..beed623 100644
> --- a/pan/ltp-pan.c
> +++ b/pan/ltp-pan.c
> @@ -645,7 +645,7 @@ int main(int argc, char **argv)
>  				strerror(errno));
>  		fprintf(logfile,
>  			"\n-----------------------------------------------\n");
> -		fprintf(logfile, "Total Tests: %d\n", coll->cnt);
> +		fprintf(logfile, "Total Tests: %d\n", (coll->cnt - starts));

This change does not make much sense to me. The coll->cnt is number of
tests loaded from the runtest file, right? The starts is basically
number of tests to be executed. The runltp script executes ltp-pan with
-S which means that starts is set to coll->cnt in the main() function
and so after this change it would print "Total Tests: 0" regardless.

-- 
Cyril Hrubis
chrubis@suse.cz

[LTP] [PATCH] pan/ltp-pan.c: fix incorrect number of total tests

[Xiao Yang]

On 2017/04/10 21:50, Cyril Hrubis wrote:
> Hi!
>>   pan/ltp-pan.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/pan/ltp-pan.c b/pan/ltp-pan.c
>> index 1614c70..beed623 100644
>> --- a/pan/ltp-pan.c
>> +++ b/pan/ltp-pan.c
>> @@ -645,7 +645,7 @@ int main(int argc, char **argv)
>>   				strerror(errno));
>>   		fprintf(logfile,
>>   			"\n-----------------------------------------------\n");
>> -		fprintf(logfile, "Total Tests: %d\n", coll->cnt);
>> +		fprintf(logfile, "Total Tests: %d\n", (coll->cnt - starts));
> This change does not make much sense to me. The coll->cnt is number of
> tests loaded from the runtest file, right? The starts is basically
> number of tests to be executed. The runltp script executes ltp-pan with
> -S which means that starts is set to coll->cnt in the main() function
> and so after this change it would print "Total Tests: 0" regardless.
>
Hi Cyril

Actually, the coll->cnt is number of tests loaded from the runtest file.
However, starts is set to coll->cnt at first , and it reduces when 
completing one test.
Please see the following code:
cpid =
run_child(coll->ary[c], running + i, quiet_mode,
&failcnt, fmt_print, logfile);
if (cpid != -1)
++num_active;
if ((cpid != -1 || sequential) && starts > 0)
--starts;

I think that ltp-pan could indicate the number of tests which have been 
run when receiving one signal.
I am not familiar with ltp-pan code，so do you have some suggestions 
about it?

Thanks,
Xiao Yang.

[LTP] [PATCH] pan/ltp-pan.c: fix incorrect number of total tests

[Cyril Hrubis]

Hi!
> Actually, the coll->cnt is number of tests loaded from the runtest file.
> However, starts is set to coll->cnt at first , and it reduces when 
> completing one test.
> Please see the following code:
> cpid =
> run_child(coll->ary[c], running + i, quiet_mode,
> &failcnt, fmt_print, logfile);
> if (cpid != -1)
> ++num_active;
> if ((cpid != -1 || sequential) && starts > 0)
> --starts;
> 
> I think that ltp-pan could indicate the number of tests which have been 
> run when receiving one signal.

Hmm, ok, so unless starts is set to -1, which means "execute the testrun
until interrupted" this would result in the number of executed tests, if
starts is -1 the result is incorrect anyway.

But to get it really correct we should rather add a counter and count
each started test child or something.

> I am not familiar with ltp-pan code???so do you have some suggestions 
> about it?

Well the ltp-pan code is messy and the tool is in a maintenance mode,
which means that we only do bugfixes. This looks like a bug so I'm not
against fixing it but at the same time I do not think that this is
important enough. So my suggestion would be to wait for the new LTP
test runner that I started to work on some time ago and keep up with
ltp-pan.c as it is until then.

-- 
Cyril Hrubis
chrubis@suse.cz

[LTP] [PATCH] pan/ltp-pan.c: fix incorrect number of total tests

[Xiao Yang]

Hi Cyril

Thanks for your explanation. :-)
Agreed.  I would be to wait for the new LTP test runner.

Thanks,
Xiao Yang

On 2017/04/12 23:31, Cyril Hrubis wrote:
> Hi!
>> Actually, the coll->cnt is number of tests loaded from the runtest file.
>> However, starts is set to coll->cnt at first , and it reduces when
>> completing one test.
>> Please see the following code:
>> cpid =
>> run_child(coll->ary[c], running + i, quiet_mode,
>> &failcnt, fmt_print, logfile);
>> if (cpid != -1)
>> ++num_active;
>> if ((cpid != -1 || sequential)&&  starts>  0)
>> --starts;
>>
>> I think that ltp-pan could indicate the number of tests which have been
>> run when receiving one signal.
> Hmm, ok, so unless starts is set to -1, which means "execute the testrun
> until interrupted" this would result in the number of executed tests, if
> starts is -1 the result is incorrect anyway.
>
> But to get it really correct we should rather add a counter and count
> each started test child or something.
>
>> I am not familiar with ltp-pan code???so do you have some suggestions
>> about it?
> Well the ltp-pan code is messy and the tool is in a maintenance mode,
> which means that we only do bugfixes. This looks like a bug so I'm not
> against fixing it but at the same time I do not think that this is
> important enough. So my suggestion would be to wait for the new LTP
> test runner that I started to work on some time ago and keep up with
> ltp-pan.c as it is until then.
>