[protobuf] Unittests failing on armel architecture
Hi,
I'm one of the Debian maintainers for protobuf and 2.3.0 fails during
unittests on the armel architecture (only on it). This might be a host
issue or a CFLAGS issue (we had something like this for sh4, where
'-mieee' is needed, for example).
In any case, here are the failues we see in the logs:
[ RUN ] CustomOptions.OptionLocations
google/protobuf/descriptor_unittest.cc:1979: Failure
Value of: service-options().GetExtension(protobuf_unittest::service_opt1)
Actual: -8589934594
Expected: -9876543210LL
Which is: -9876543210
[ FAILED ] CustomOptions.OptionLocations (4 ms)
[ RUN ] CustomOptions.OptionTypes
google/protobuf/descriptor_unittest.cc:1999: Failure
Value of: options-GetExtension(protobuf_unittest::sint64_opt)
Actual: -9223372035781033984
Expected: kint64min
Which is: -9223372036854775808
google/protobuf/descriptor_unittest.cc:2013: Failure
Value of: options-GetExtension(protobuf_unittest::sint64_opt)
Actual: 9223372035781033983
Expected: kint64max
Which is: 9223372036854775807
[ FAILED ] CustomOptions.OptionTypes (2 ms)
[ RUN ] ExtensionSetTest.Parsing
google/protobuf/test_util.cc:1127: Failure
Value of: message.GetExtension(unittest::optional_sint64_extension )
Actual: 0
Expected: 106
google/protobuf/test_util.cc:1184: Failure
Value of: message.GetExtension(unittest::repeated_sint64_extension , 0)
Actual: 0
Expected: 206
google/protobuf/test_util.cc:1212: Failure
Value of: message.GetExtension(unittest::repeated_sint64_extension , 1)
Actual: 0
Expected: 306
google/protobuf/test_util.cc:1265: Failure
Value of: message.GetExtension(unittest::default_sint64_extension )
Actual: 0
Expected: 406
[ FAILED ] ExtensionSetTest.Parsing (4 ms)
There are also some other tests failing, but this is probably because
some dependencies are missing:
libprotobuf FATAL google/protobuf/test_util_lite.cc:950] CHECK failed: (106) ==
(message.GetExt
/bin/bash: line 4: 26978 Aborted ${dir}$tst
FAIL: protobuf-lite-test
Can you confirm that we need 'unzip' and 'jar' to be available during
unittests? Or why does protobuf-lite-test 'FAIL' hard (instead of
exiting with no error if these are just optional deps?)
In any case, the full build log is at
https://buildd.debian.org/fetch.cgi?pkg=protobufarch=armelver=2.3.0-1stamp=1264814475file=log,
if you're interested.
Thanks in advance,
iustin
--
You received this message because you are subscribed to the Google Groups
Protocol Buffers group.
To post to this group, send email to proto...@googlegroups.com.
To unsubscribe from this group, send email to
protobuf+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/protobuf?hl=en.
Re: [protobuf] Unittests failing on armel architecture
On Mon, Feb 1, 2010 at 1:22 PM, Iustin Pop iu...@k1024.org wrote:
Hi,
I'm one of the Debian maintainers for protobuf and 2.3.0 fails during
unittests on the armel architecture (only on it). This might be a host
issue or a CFLAGS issue (we had something like this for sh4, where
'-mieee' is needed, for example).
In any case, here are the failues we see in the logs:
[ RUN ] CustomOptions.OptionLocations
google/protobuf/descriptor_unittest.cc:1979: Failure
Value of: service-options().GetExtension(protobuf_unittest::service_opt1)
Actual: -8589934594
Expected: -9876543210LL
Which is: -9876543210
[ FAILED ] CustomOptions.OptionLocations (4 ms)
[ RUN ] CustomOptions.OptionTypes
google/protobuf/descriptor_unittest.cc:1999: Failure
Value of: options-GetExtension(protobuf_unittest::sint64_opt)
Actual: -9223372035781033984
Expected: kint64min
Which is: -9223372036854775808
google/protobuf/descriptor_unittest.cc:2013: Failure
Value of: options-GetExtension(protobuf_unittest::sint64_opt)
Actual: 9223372035781033983
Expected: kint64max
Which is: 9223372036854775807
[ FAILED ] CustomOptions.OptionTypes (2 ms)
$ python
%x % 9223372036854775808
'8000'
%x % 9223372035781033984
'7fffc000'
So it looks like somehow the bits are getting cut off. I suspect a compiler
bug of some sort. Perhaps the auto-generated copy construct for struct
ExtensionSet::Extension is not copying the union correctly? Unfortunately I
can't really help debug as I don't have access to this platform.
[ RUN ] ExtensionSetTest.Parsing
google/protobuf/test_util.cc:1127: Failure
Value of: message.GetExtension(unittest::optional_sint64_extension )
Actual: 0
Expected: 106
google/protobuf/test_util.cc:1184: Failure
Value of: message.GetExtension(unittest::repeated_sint64_extension , 0)
Actual: 0
Expected: 206
google/protobuf/test_util.cc:1212: Failure
Value of: message.GetExtension(unittest::repeated_sint64_extension , 1)
Actual: 0
Expected: 306
google/protobuf/test_util.cc:1265: Failure
Value of: message.GetExtension(unittest::default_sint64_extension )
Actual: 0
Expected: 406
[ FAILED ] ExtensionSetTest.Parsing (4 ms)
Same bug, I think. The lower 30 bits are being zeroed, which completely
clobbers these small numbers.
There are also some other tests failing, but this is probably because
some dependencies are missing:
libprotobuf FATAL google/protobuf/test_util_lite.cc:950] CHECK failed:
(106) == (message.GetExt
/bin/bash: line 4: 26978 Aborted ${dir}$tst
FAIL: protobuf-lite-test
This again looks like the same bug, but this time in protobuf-lite-test,
whose output is not as nice as protobuf-test as it doesn't use the Google
Test framework.
Can you confirm that we need 'unzip' and 'jar' to be available during
unittests? Or why does protobuf-lite-test 'FAIL' hard (instead of
exiting with no error if these are just optional deps?)
protobuf-lite-test does not require these. The error messages about these
are being printed by zip_output_unittest, which is run after
protobuf-lite-test. zip_output_unittest is designed to skip the tests that
use unzip and jar if the tools are not available, which is why the build log
says it passed despite the errors. protobuf-lite-test is failing for other
reasons.
--
You received this message because you are subscribed to the Google Groups
Protocol Buffers group.
To post to this group, send email to proto...@googlegroups.com.
To unsubscribe from this group, send email to
protobuf+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/protobuf?hl=en.
Re: [protobuf] Unittests failing on armel architecture
On Mon, Feb 01, 2010 at 02:42:53PM -0800, Kenton Varda wrote:
On Mon, Feb 1, 2010 at 1:22 PM, Iustin Pop iu...@k1024.org wrote:
Hi,
I'm one of the Debian maintainers for protobuf and 2.3.0 fails during
unittests on the armel architecture (only on it). This might be a host
issue or a CFLAGS issue (we had something like this for sh4, where
'-mieee' is needed, for example).
In any case, here are the failues we see in the logs:
[ RUN ] CustomOptions.OptionLocations
google/protobuf/descriptor_unittest.cc:1979: Failure
Value of: service-options().GetExtension(protobuf_unittest::service_opt1)
Actual: -8589934594
Expected: -9876543210LL
Which is: -9876543210
[ FAILED ] CustomOptions.OptionLocations (4 ms)
[ RUN ] CustomOptions.OptionTypes
google/protobuf/descriptor_unittest.cc:1999: Failure
Value of: options-GetExtension(protobuf_unittest::sint64_opt)
Actual: -9223372035781033984
Expected: kint64min
Which is: -9223372036854775808
google/protobuf/descriptor_unittest.cc:2013: Failure
Value of: options-GetExtension(protobuf_unittest::sint64_opt)
Actual: 9223372035781033983
Expected: kint64max
Which is: 9223372036854775807
[ FAILED ] CustomOptions.OptionTypes (2 ms)
$ python
%x % 9223372036854775808
'8000'
%x % 9223372035781033984
'7fffc000'
So it looks like somehow the bits are getting cut off. I suspect a compiler
bug of some sort. Perhaps the auto-generated copy construct for struct
ExtensionSet::Extension is not copying the union correctly? Unfortunately I
can't really help debug as I don't have access to this platform.
Thanks for the hint. Will see what we can do about it (at least check if there
are no known issues with gcc on that platform).
[ RUN ] ExtensionSetTest.Parsing
google/protobuf/test_util.cc:1127: Failure
Value of: message.GetExtension(unittest::optional_sint64_extension )
Actual: 0
Expected: 106
google/protobuf/test_util.cc:1184: Failure
Value of: message.GetExtension(unittest::repeated_sint64_extension , 0)
Actual: 0
Expected: 206
google/protobuf/test_util.cc:1212: Failure
Value of: message.GetExtension(unittest::repeated_sint64_extension , 1)
Actual: 0
Expected: 306
google/protobuf/test_util.cc:1265: Failure
Value of: message.GetExtension(unittest::default_sint64_extension )
Actual: 0
Expected: 406
[ FAILED ] ExtensionSetTest.Parsing (4 ms)
Same bug, I think. The lower 30 bits are being zeroed, which completely
clobbers these small numbers.
ACK.
There are also some other tests failing, but this is probably because
some dependencies are missing:
libprotobuf FATAL google/protobuf/test_util_lite.cc:950] CHECK failed:
(106) == (message.GetExt
/bin/bash: line 4: 26978 Aborted ${dir}$tst
FAIL: protobuf-lite-test
This again looks like the same bug, but this time in protobuf-lite-test,
whose output is not as nice as protobuf-test as it doesn't use the Google
Test framework.
ACK.
Can you confirm that we need 'unzip' and 'jar' to be available during
unittests? Or why does protobuf-lite-test 'FAIL' hard (instead of
exiting with no error if these are just optional deps?)
protobuf-lite-test does not require these. The error messages about these
are being printed by zip_output_unittest, which is run after
protobuf-lite-test. zip_output_unittest is designed to skip the tests that
use unzip and jar if the tools are not available, which is why the build log
says it passed despite the errors. protobuf-lite-test is failing for other
reasons.
Ah OK, I see, sorry for misreading the logs.
thanks for debugging this!
iustin
--
You received this message because you are subscribed to the Google Groups
Protocol Buffers group.
To post to this group, send email to proto...@googlegroups.com.
To unsubscribe from this group, send email to
protobuf+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/protobuf?hl=en.
Re: [protobuf] Unittests failing on armel architecture
On Mon, Feb 1, 2010 at 2:42 PM, Kenton Varda ken...@google.com wrote: On Mon, Feb 1, 2010 at 1:22 PM, Iustin Pop iu...@k1024.org wrote: Hi, I'm one of the Debian maintainers for protobuf and 2.3.0 fails during unittests on the armel architecture (only on it). This might be a host issue or a CFLAGS issue (we had something like this for sh4, where '-mieee' is needed, for example). In any case, here are the failues we see in the logs: [ RUN ] CustomOptions.OptionLocations google/protobuf/descriptor_unittest.cc:1979: Failure Value of: service-options().GetExtension(protobuf_unittest::service_opt1) Actual: -8589934594 Expected: -9876543210LL Which is: -9876543210 [ FAILED ] CustomOptions.OptionLocations (4 ms) [ RUN ] CustomOptions.OptionTypes google/protobuf/descriptor_unittest.cc:1999: Failure Value of: options-GetExtension(protobuf_unittest::sint64_opt) Actual: -9223372035781033984 Expected: kint64min Which is: -9223372036854775808 google/protobuf/descriptor_unittest.cc:2013: Failure Value of: options-GetExtension(protobuf_unittest::sint64_opt) Actual: 9223372035781033983 Expected: kint64max Which is: 9223372036854775807 [ FAILED ] CustomOptions.OptionTypes (2 ms) $ python %x % 9223372036854775808 '8000' %x % 9223372035781033984 '7fffc000' Oops, I used the wrong numbers here. The interesting numbers are: print \n%x != %x\n%x != %x % tuple((x 0x for x in (-9223372035781033984, -9223372036854775808, 9223372035781033983, 9223372036854775807))) 80004000 != 8000 7fffbfff != 7fff So it looks like what we're actually seeing here is one bit of corruption. I still strongly suspect a compiler bug. Probably the compiler's emulation of 64-bit arithmetic is broken in some way. This reminds me of another bug report we had awhile back, but I don't think they are exactly the same bug (look at message #19 in the thread): http://groups.google.com/group/protobuf/browse_frm/thread/e2cb856aad6c81f6/b32e3460f10af3aa -- You received this message because you are subscribed to the Google Groups Protocol Buffers group. To post to this group, send email to proto...@googlegroups.com. To unsubscribe from this group, send email to protobuf+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/protobuf?hl=en.
