[lfs-dev] glibc-2.23

ALZ (phyglos.org) alz at phyglos.org
Sun Feb 21 12:35:05 PST 2016


On 02/20/2016 09:49 PM, Bruce Dubbs wrote:
> Bruce Dubbs wrote:
>> Douglas R. Reno wrote:
>>> On Fri, Feb 19, 2016 at 3:18 PM, Bruce Dubbs <bruce.dubbs at gmail.com>
>>> wrote:
>>>
>>>> I've built the new glibc in my sandbox and will start doing a -rc2
>>>> when my
>>>> full build completes in the next hour or so.
>>>>
>>>> I did look at the test failures:
>>>>
>>>> XPASS: elf/tst-protected1a
>>>> XPASS: elf/tst-protected1b
>>>> FAIL: posix/tst-getaddrinfo4
>>>> FAIL: posix/tst-getaddrinfo5
>>>> Summary of test results:
>>>>        2 FAIL
>>>>     2401 PASS
>>>>       84 XFAIL
>>>>        2 XPASS
>>>>
>>>> I've updated the text to add posix/tst-getaddrinfo5 to the list of
>>>> known
>>>> failures.  When I look at the text we have now, I also see:
>>>>
>>>> * The rt/tst-cputimer1 and rt/tst-cpuclock2 tests have been known to
>>>> fail.
>>>> The reason is not completely understood, but indications are that minor
>>>> timing issues can trigger these failures.
>>>>
>>>> * The math tests sometimes fail when running on systems where the
>>>> CPU is
>>>> not a relatively new Intel or AMD processor.
>>>>
>>>> * Other tests known to fail on some architectures are
>>>> malloc/tst-malloc-usable and nptl/tst-cleanupx4.
>>>>
>>>> I have already removed the text about tst-protected1{a,b}.
>>>>
>>>> I have not seen any of these in a long time. Should I remove them?
>>
>>> Are these i686 specific?
>>
>> I don't think so, but I'm not sure.  I can do a build on my 686 and
>> check,
>> but that wouldn't hold off proceeding with BLFS testing.  I'll try to set
>> it up tonight and let it run to check.  A full build with all tests takes
>> about 17 hours on that system.
>
> Well it took 18 hours, but the 7.9-rc2 built on my 686.  For glibc, I
> had one additional test failure: nptl/tst-cleanupx4.  That is mentioned
> above.

Over here, after +3h for Ch5 and +17h for Ch6, on a Pentium M 1.6 Ghz 
(i686) with 2GB RAM.

Here the results of glibc-2.23:

XPASS: elf/tst-protected1a
XPASS: elf/tst-protected1b
FAIL: nptl/tst-cleanupx4
FAIL: posix/tst-getaddrinfo4
FAIL: posix/tst-getaddrinfo5
FAIL: stdio-common/test-vfprintf
Summary of test results:
       4 FAIL
    2377 PASS
      84 XFAIL
       2 XPASS

On a Q9950 @3.4Ghz with 12 GB RAM (x86_64), expected results:

XPASS: elf/tst-protected1a
XPASS: elf/tst-protected1b
FAIL: posix/tst-getaddrinfo4
FAIL: posix/tst-getaddrinfo5
Summary of test results:
       2 FAIL
    2401 PASS
      84 XFAIL
       2 XPASS

In a VMware guest (x86_64) with 4 GB, interesting combination:

XPASS: conform/UNIX98/ndbm.h/linknamespace
XPASS: conform/XOPEN2K/ndbm.h/linknamespace
XPASS: conform/XOPEN2K8/ndbm.h/linknamespace
XPASS: conform/XPG4/ndbm.h/linknamespace
XPASS: elf/tst-protected1a
XPASS: elf/tst-protected1b
FAIL: stdio-common/test-vfprintf
Summary of test results:
       1 FAIL
    2402 PASS
      80 XFAIL
       6 XPASS

>
> I also had a bunch of gcc test failures not in x86-64.  There were 8
> failures using different options with c-c++-common/asan/null-deref-1.c,
> 1 failure for gcc.dg/pr45352-1.c, and 1 for gcc.dg/pr63914.c. The test
> failures that are in the x86_64 build, directory_iterator.cc and
> recursive_directory_iterator.cc fail here too.
>
> The only other failure I see that differs from x86-64 is:
>
> 105-inetutils-1.9.4:FAIL: ping-localhost.sh
>

On the physical x86_64, clean results:

		=== g++ Summary ===
# of expected passes		93742
# of unexpected successes	2
# of expected failures		342
# of unsupported tests		3746

		=== gcc Summary ===
# of expected passes		114783
# of expected failures		262
# of unsupported tests		1806

		=== libatomic Summary ===
# of expected passes		54

		=== libgomp Summary ===

# of expected passes		1617
# of unsupported tests		170

		=== libitm Summary ===
# of expected passes		26
# of expected failures		3
# of unsupported tests		1

		=== libstdc++ Summary ===
# of expected passes		9871
# of unexpected failures	2
# of expected failures		65
# of unsupported tests		530

The 2 unexpected failures are the same that appear on virtual x86_64:

FAIL: experimental/filesystem/iterators/recursive_directory_iterator.cc 
execution test
FAIL: experimental/filesystem/iterators/directory_iterator.cc execution test

On the i686 laptop, several timeouts and up to 12 unexpected failures...


With regards 'segfaults' and 'traps' in /var/log/kern.log while testing gcc:

* on i686 there are 216 segfaults, no traps
* on physical x86_64, 138 segfaults plus 57 traps
* on virtual x86_64, 163 segfaults and 73 traps.

Hope it helps.
Alz






More information about the lfs-dev mailing list