2019-10-21 09:15:47 UTC - Sujith Ramanathan: @Sujith Ramanathan has joined the
channel
----
2019-10-21 12:48:58 UTC - Ivan Kelly: @Retardust for the most part it's
pointless, but some people like to keep it for consistency. It is needed in
some cases where the values used do cost something to compute, but these are
rare
----
2019-10-21 12:59:51 UTC - Matteo Merli: The reason for that is that
`log.debug(“.... {}”, x)` gets actually converted into `log.debug(“.... {}”,
new Object[] {x})`.
The array object gets created regardless of whether the debug is printed.
----
2019-10-21 13:00:47 UTC - Ivan Kelly: doesn't escape analyse handle the new
Object[]?
----
2019-10-21 13:02:54 UTC - Matteo Merli: Escape analysis is a bit hit and miss
and from heap dumps doesn’t look it work for this case.
My impression is that it gets disabled if you pass a ref to a method. Only
works if you return a ref from a method and it gets used locally.
----
2019-10-21 13:42:34 UTC - Ron Wheeler: @Ron Wheeler has joined the channel
----
2019-10-21 15:51:07 UTC - Retardust: Do you test that with jmh?
----
2019-10-21 16:02:57 UTC - Matteo Merli: yes, some time back, with `-prof gc`
----
2019-10-21 16:55:04 UTC - tuteng: It seems that the version of pyyaml is too
low, resulting in the FullLoader attribute not being found.
python-yaml=>3.13 python3-yaml=>3.13, version with attribute FullLoader
=> 5.1.2
----
2019-10-21 19:03:10 UTC - Addison Higham: this still an issue? was just going
to retrigger some tests, but maybe I should hold off?
----
2019-10-22 05:27:18 UTC - Jerry Peng: the temporary fix i.e. reverting a PR is
merged
----
