macOS crashes with 1.69.0 when using conda packages on Xcode >= 7 about boost-cpp-feedstock HOT 9 CLOSED

It does look like 1.69.0 is installed, but it’s also detecting 1.67.0. Not entirely sure what is going on, but if it’s mixing two different Boost installs that could cause this problem. Is it possible it’s picking Boost up from Homebrew or elsewhere?

from boost-cpp-feedstock.

Note: these crashes aren't present in Xcode 6.4

from boost-cpp-feedstock.

Apologies, it seems the problem is present on older versions of boost-cpp

https://travis-ci.org/wesm/arrow/jobs/469307447#L1809

I'll do some searching to see if it's a known issue

from boost-cpp-feedstock.

That "1.67.0" output is some build system output generated by

https://github.com/apache/arrow/blob/master/cpp/cmake_modules/ThirdpartyToolchain.cmake#L170

It's not actually finding 1.67.0 on the system.

I was able to reproduce the crash with 1.69.0 packages but passing build with 1.68.0 on Xcode 8.3:

• Passing build with 1.68.0 pinned https://travis-ci.org/apache/arrow/jobs/469742389
• Failing build with 1.69.0 https://travis-ci.org/apache/arrow/jobs/469666491#L2203

I'm working around the problem for now in our CI by pinning 1.68.0, but this is somewhat of a ticking time bomb because we have boost-cpp in both build and run scope in arrow-cpp

https://github.com/conda-forge/arrow-cpp-feedstock/blob/master/recipe/meta.yaml#L30

cc @kszucs @xhochy @pitrou

from boost-cpp-feedstock.

It's not actually finding 1.67.0 on the system.

No. See https://travis-ci.org/wesm/arrow/jobs/469307447#L2269-L2273 and https://travis-ci.org/wesm/arrow/jobs/469307447#L2280-L2282

from boost-cpp-feedstock.

That's a different issue. I think @jakirkham was talking about

https://travis-ci.org/wesm/arrow/jobs/469307447#L852

I'm not sure why it is loading the dylibs from /usr/local/lib because we prepended the conda lib path to LD_LIBRARY_PATH.

In the failures I linked above Boost is being statically linked so this /usr/local/lib issue isn't relevant

from boost-cpp-feedstock.

Can you show me the exact line of the failure you are talking about?

from boost-cpp-feedstock.

Yes, this segmentation fault:

https://travis-ci.org/apache/arrow/jobs/469666491#L2203

Two other developers have confirmed the problem with 1.69.0 locally but that 1.68.0 works as expected:

• @kszucs https://issues.apache.org/jira/browse/ARROW-4050?focusedCommentId=16724164&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-16724164
• @tanyaschlusser https://issues.apache.org/jira/browse/ARROW-4050?focusedCommentId=16723180&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-16723180

from boost-cpp-feedstock.

1.69 is not the current version anymore

from boost-cpp-feedstock.