Test suite fails on macOS after update to 9.6.13 and 11.3
The postgresql updates to 9.6.13 and 11.3 ends up breaking tests on my macOS machine. The log is attached, but it looks like a serious issue that may require significant changes. Here is the full build log for 11.3:
https://nix-cache.s3.amazonaws.com/log/8ip66iqpchrqq77l7f4ccrrks5zh1bav-postgresql-11.3.drv <https://nix-cache.s3.amazonaws.com/log/8ip66iqpchrqq77l7f4ccrrks5zh1bav-postgresql-11.3.drv>
I suspect this is caused by 808e1e75fb376ff7b39d0c3712fa85ff6e0f0720 <https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=808e1e75fb376ff7b39d0c3712fa85ff6e0f0720> and /messages/by-id/20190408064141.GA2016666@rfd.leadboat.com </messages/by-id/20190408064141.GA2016666@rfd.leadboat.com>.
Attachments:
On 2019-Jun-06, Matthew Bauer wrote:
The postgresql updates to 9.6.13 and 11.3 ends up breaking tests on my macOS machine. The log is attached, but it looks like a serious issue that may require significant changes. Here is the full build log for 11.3:
https://nix-cache.s3.amazonaws.com/log/8ip66iqpchrqq77l7f4ccrrks5zh1bav-postgresql-11.3.drv <https://nix-cache.s3.amazonaws.com/log/8ip66iqpchrqq77l7f4ccrrks5zh1bav-postgresql-11.3.drv>
Hmm ... it seems it's psql that is crashing, not the server, right? Can
you grab a backtrace from those core dumps? Does it still crash without
your less-is-more patch?
--
�lvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
Matthew Bauer <mjbauer95@gmail.com> writes:
The postgresql updates to 9.6.13 and 11.3 ends up breaking tests on my macOS machine. The log is attached, but it looks like a serious issue that may require significant changes. Here is the full build log for 11.3:
https://nix-cache.s3.amazonaws.com/log/8ip66iqpchrqq77l7f4ccrrks5zh1bav-postgresql-11.3.drv <https://nix-cache.s3.amazonaws.com/log/8ip66iqpchrqq77l7f4ccrrks5zh1bav-postgresql-11.3.drv>
It's been awhile since I tried to work on a Mac that has System Integrity
Protection turned on, but I think that SIP causes "make check" to fail
in more or less this way, because (for no defensible reason) it strips
DYLD_LIBRARY_PATH from the environment that's passed down to psql,
causing failure to find the right libpq.dylib file.
If you don't wish to turn off SIP, a workaround is to do "make install"
before "make check".
regards, tom lane