This page is not created by, affiliated with, or supported by Slack Technologies, Inc.
2018-09-09
Channels
- # announcements (1)
- # beginners (166)
- # calva (2)
- # cider (7)
- # clojure (21)
- # clojure-dev (11)
- # clojure-italy (3)
- # clojure-russia (5)
- # clojure-spec (13)
- # clojure-uk (1)
- # clojurescript (251)
- # core-async (2)
- # cursive (14)
- # emacs (6)
- # events (8)
- # figwheel-main (5)
- # fulcro (17)
- # off-topic (8)
- # pedestal (3)
- # tools-deps (17)
https://clojurians.slack.com/archives/C06E3HYPR/p1536440903000100 It totally has the memory leak feature @gfredericks https://dev.clojure.org/jira/browse/CLJ-1152
oh the cache is at the protocol level? I imagined it was at the call site
in which case you'd have to work just slightly harder to force the leak
but yeah okay that's cool
That bug is specifically in eval though so it’s a little subtle
that bug ain't in eval - it's the lack of a bound unconditional retention within the inline cache
would pay to see a superhero movie called Eval the Exacerbater
So would it be at least semi-accurate to say that the places where Var re-assignment works in tools.trace are probably the same places where it works with spec's instrument, and they both fail in the same kinds of calls, too, which I think are "protocol method calls that do not go through a Var, and direct linked calls"?
Yes, same technique. Also inlined calls. And primitive type hinted functions (not sure if those have same issues in tools.trace).