15:06 <maurer> _habnabit: If you don't care too much about package manager hygeine, you can use the program "alien" to convert the .deb to a .rpm

15:07 <maurer> _habnabit: it may not be linked perfectly, but libc hasn't updated in a while, so there's a chance it'll be fine

15:50 <_habnabit> maurer, ah, i ended up just using nix

15:53 <maurer> Huh, I'm surprised nix works cleanly on centos.

17:39 <isd> So, node is currently using up ~650MiB of ram, and sandstorm is failing to actually fill-in an offer iframe.

17:46 <isd> A while back it went haywire and actually *ran out* of ram, couldn't spawn grains even. I'd already power-cycled it by then, but someone gave me a js command to take a heap snapshot. Having a hard time digging that up again.

19:10 <zarvox> isd: Meteor.call("heapdump")

19:12 <zarvox> isd: if your user is a server admin, that should make node write a heapdump to /opt/sandstorm/var/log/$TIMESTAMP.heapsnapshot

19:40 <_habnabit> maurer, i was too, kinda

19:41 <_habnabit> does capnp have a way to make sure your .capnp changes are backward-compatible?

19:43 <maurer> _habnabit: Don't change the numbers after an @ sign, don't change the type of a field, and that's about it. Don't think there's an explicit "is this an extension of this" checker though

19:44 <_habnabit> maurer, well sure i know how but i like static analysis

19:45 <mokomull> hm, you'd need something with both versions of the source file available, right? like some hairy integration with your favorite VCS

19:46 <_habnabit> doesn't need to be that hairy

19:47 <_habnabit> the usual mode of operation for tools like this is to take the current file and a diff

19:47 <_habnabit> you can reconstruct the previous version from that and kinda figure out what parts changed

19:48 <_habnabit> ofc parsing a diff to figure out what nodes in an AST were affected is.. hard

19:51 <maurer> I don't know that the tool exists. If it doesn't, the ideal way to write it would be to invoke capnpc to spit out a capnproto encoded version of each schema (to ditch anything related to the text) and then read those in and reason about whether the nodes were compatible

19:52 <_habnabit> maybe i'll do that