Test failures on OpenBSD 6.4

[dveeden]

http://www.cpantesters.org/cpan/report/16a30d94-22ba-11e9-b2ea-9ec3a8a2fe4a Might be a good opportunity to look at #120

[pali]

This looks like a same issue as for DBD::MariaDB from comment: https://github.com/gooddata/DBD-MariaDB/issues/80#issuecomment-404682682 (Temporary) fix for DBD::MariaDB: https://github.com/gooddata/DBD-MariaDB/commit/77521260a8e6c987e920f425b4c5dce4a6403456 It seems that OpenBSD malloc overhead is bigger then on Linux. Some more debugging is needed.

[dveeden]

Note that p5-Proc-ProcessTable-0.53p1 results in

t/60leaks.t ............................. skipped: module Proc::ProcessTable does not support size attribute on current platform

So this needs a newer version of Proc::ProcessTable (the test report says 0.55)

[dveeden]

OpenBSD has MALLOC_OPTIONS, but using that to enable/disable features doesn't seem to make a real difference.

[dveeden]

This seems to happen with both clang and gcc (as expected)

[dveeden]

https://www.openbsd.org/papers/eurobsdcon2009/otto-malloc.pdf https://www.openbsd.org/innovations.html "malloc(3) randomization implemented by Thierry Deval. Guard pages and randomized (delayed) free added by Ted Unangst. Reimplemented by Otto Moerbeek for OpenBSD 4.4. "

[dveeden]

The plan here is to re-write the test.