[LTP] syscalls/mallinfo01 test is broken

[LTP] syscalls/mallinfo01 test is broken

[Martin Doucha]

Hello,
the mallinfo01 test is failing on some of our test machines because it
makes wrong assumptions about how malloc() and free() will change the
number of free chunks (ordblks). I recommend we remove the test from
runfiles before the upcoming release and fix it later because part of
the test scenario needs major redesign.

> tst_test.c:1313: TINFO: Timeout per run is 0h 18m 00s
> mallinfo_common.h:16: TINFO: Start...
> mallinfo_common.h:18: TINFO: arena: 135168
> mallinfo_common.h:19: TINFO: ordblks: 1
> mallinfo_common.h:20: TINFO: smblks: 1
> mallinfo_common.h:21: TINFO: hblks: 0
> mallinfo_common.h:22: TINFO: hblkhd: 0
> mallinfo_common.h:23: TINFO: usmblks: 0
> mallinfo_common.h:24: TINFO: fsmblks: 64
> mallinfo_common.h:25: TINFO: uordblks: 128
> mallinfo_common.h:26: TINFO: fordblks: 135040
> mallinfo_common.h:27: TINFO: keepcost: 134976
> mallinfo_common.h:16: TINFO: Test uordblks...
> mallinfo_common.h:18: TINFO: arena: 135168
> mallinfo_common.h:19: TINFO: ordblks: 2
> mallinfo_common.h:20: TINFO: smblks: 0
> mallinfo_common.h:21: TINFO: hblks: 0
> mallinfo_common.h:22: TINFO: hblkhd: 0
> mallinfo_common.h:23: TINFO: usmblks: 0
> mallinfo_common.h:24: TINFO: fsmblks: 0
> mallinfo_common.h:25: TINFO: uordblks: 34048
> mallinfo_common.h:26: TINFO: fordblks: 101120
> mallinfo_common.h:27: TINFO: keepcost: 101056
> mallinfo01.c:46: TPASS: mallinfo() uordblks passed
> mallinfo_common.h:16: TINFO: Test ordblks...
> mallinfo_common.h:18: TINFO: arena: 135168
> mallinfo_common.h:19: TINFO: ordblks: 3
> mallinfo_common.h:20: TINFO: smblks: 0
> mallinfo_common.h:21: TINFO: hblks: 0
> mallinfo_common.h:22: TINFO: hblkhd: 0
> mallinfo_common.h:23: TINFO: usmblks: 0
> mallinfo_common.h:24: TINFO: fsmblks: 0
> mallinfo_common.h:25: TINFO: uordblks: 32432
> mallinfo_common.h:26: TINFO: fordblks: 102736
> mallinfo_common.h:27: TINFO: keepcost: 101056
> mallinfo01.c:58: TFAIL: mallinfo() ordblks failed
> 
> Summary:
> passed   1
> failed   1
> broken   0
> skipped  0
> warnings 0

-- 
Martin Doucha   mdoucha@suse.cz
QA Engineer for Software Maintenance
SUSE LINUX, s.r.o.
CORSO IIa
Krizikova 148/34
186 00 Prague 8
Czech Republic

[LTP] syscalls/mallinfo01 test is broken

[Cyril Hrubis]

Hi!
> the mallinfo01 test is failing on some of our test machines because it
> makes wrong assumptions about how malloc() and free() will change the
> number of free chunks (ordblks). I recommend we remove the test from
> runfiles before the upcoming release and fix it later because part of
> the test scenario needs major redesign.

I've started some pre-release testruns and I've seen some failures as
well. I guess that removing it from runtest file and fixing it after
release sounds sane at this point.

-- 
Cyril Hrubis
chrubis@suse.cz

[LTP] syscalls/mallinfo01 test is broken

[xuyang2018.jy]

Hi Cyril, Martin
> Hi!
>> the mallinfo01 test is failing on some of our test machines because it
>> makes wrong assumptions about how malloc() and free() will change the
>> number of free chunks (ordblks). I recommend we remove the test from
>> runfiles before the upcoming release and fix it later because part of
>> the test scenario needs major redesign.
>
> I've started some pre-release testruns and I've seen some failures as
> well. I guess that removing it from runtest file and fixing it after
> release sounds sane at this point.
This case fails because malloc uses fastbin blocks.
So the fix has two ways
1) disable fastbin by using mallopt(M_MXFAST, 0)
2) use another mallinfo struct info3 to get free chunks, so compare 
info3 and info2 ordblks member.

Best Regards
Yang Xu

>

[LTP] syscalls/mallinfo01 test is broken

[Cyril Hrubis]

Hi!
> >> the mallinfo01 test is failing on some of our test machines because it
> >> makes wrong assumptions about how malloc() and free() will change the
> >> number of free chunks (ordblks). I recommend we remove the test from
> >> runfiles before the upcoming release and fix it later because part of
> >> the test scenario needs major redesign.
> >
> > I've started some pre-release testruns and I've seen some failures as
> > well. I guess that removing it from runtest file and fixing it after
> > release sounds sane at this point.
> This case fails because malloc uses fastbin blocks.
> So the fix has two ways
> 1) disable fastbin by using mallopt(M_MXFAST, 0)
> 2) use another mallinfo struct info3 to get free chunks, so compare 
> info3 and info2 ordblks member.

If you are going to fix this test please do so as fast as possible so
that we have enough time to test the fixed test.

-- 
Cyril Hrubis
chrubis@suse.cz