On Tue, Oct 20, 2020 at 05:44:18PM +1100, Stephen Gregoratto wrote: > On 2020-10-19 10:05, Sebastien Marie wrote: > > Could you take maintainership of the port ? > > > > Personally, I am a bit reluctant to import a rust port if nobody want to > > maintain it, as it means I will need to do the work when it will break > > due to compiler update. > > Sure, I can do that. > > > From port point of vue, I only noted that it is using jemalloc-sys > > crate and so will build a local copy of jemalloc. I am unsure of the > > usage (if jemalloc is here only as dependency but not used or actively > > used), and if it will works or not. > > I'm not sure what you mean here? It's possible to patch out the use of > jemalloc at compile time (it's disabled for other OSs too). Is there > some issue with jemalloc and OpenBSD?
OpenBSD is an untested jemalloc target. I reminder to have worked a bit on it (it was a requirement/recommanded at time for rustc). Regarding building a local copy, we tend to avoid that, and prefer to use a system version. But we don't have jemalloc port, and just for one port it is fine. It permits to avoid rebuild several times the same code, centralize patches in one place, permit to follow exact version used. If it works in your tests, it is fine like that (without explictly patching it out: no patch is better). For testing if it is actively used, you could: $ make patch # extract all files $ rm -rf -- $(make show=MODCARGO_VENDOR_DIR)/jemalloc-sys/jemalloc $ make it will remove the jemalloc source directory from the crate jemalloc-sys, and build. if the port effectively uses it, it will fail to build. I will try to test it too. Thanks. -- Sebastien Marie
