Test failures with boost 1.63 package. about libbitcoin-database HOT 11 CLOSED

Yes, boost I tried making a boost 1.63.0 RPM w/ no patches, also had same failure.

I am presently working on trying to get boost 1.58.0 packaged for test environment - if that doesn't work I'll go down to 1.57.0 (what the install script built)

If that doesn't work then it is a boost packaging issue.

If boost 1.58.0 does work I'll bump up one at a time until I find the version of boost where the issue starts.

I'll figure it out. But it works with install.sh using boost 1.57.0 from source

from libbitcoin-database.

Okay just like to confirm that using boost 1.58.0 did work. I need to clean up that packaging of boost a little before I start walking up the boost versions, but its packaged the same way as the 1.63.0 that failed so I suspect there is a changed behavior in boost. I'll find the version where it breaks so maybe the changelog for that version can shed some light.

from libbitcoin-database.

Thanks, please post test-suite.log which should have specifics. Also note that the test suite successfully executed in Travis.

from libbitcoin-database.

Also, please perform a build of libbitcoin-database with its install.sh, using a non-sudo --prefix build and setting the --build-boost option. This will isolate the build as much as possible from the environment and will execute the database tests.

from libbitcoin-database.

That's test.log (test-suite.log just has the test suite summary) - I'll try the install.sh tonight and post results.

test.log.txt

from libbitcoin-database.

The failing tests:

test/hash_table.cpp(265): fatal error: in "hash_table_tests/record_hash_table__32bit__test": critical check header.read(0) == 1u has failed [2 != 1]
test/hash_table.cpp(387): fatal error: in "hash_table_tests/record_hash_table__64bit__test": critical check header.read(0) == header.empty has failed [0 != 4294967295]

Thanks, is this is little endian platform?

from libbitcoin-database.

Yes it is little endian - Xeon E3

just ran install.sh w/o build boost and got the same error.

However with the --build-boost option it passes, so the difference is boost - it doesn't like how my boost was built.

The boost I have is built on my system but using a src.rpm from Fedora Rawhide so I'll have to look at their options and patches and see if I can figure out which one is causing the problem.

from libbitcoin-database.

Ok, great - thanks for the info.

That's a bit of a relief, because the error made no sense to me at all. Not all of the libraries are yet fully verified on other than LE platforms (though the base library is). So I figured we might have an issue in the store, but sort of grasping there.

It's not obvious to me how boost could be affecting it. Please keep us posted.

from libbitcoin-database.

Looks like the test failures are the same on Arch with boost 1.63.0: http://ix.io/oQm

from libbitcoin-database.

Another boost issue, though doesn't look related:

libbitcoin/libbitcoin-system#733

from libbitcoin-database.

Closing as complete.

from libbitcoin-database.