| Geth | MoarVM: MasterDuke17++ created pull request #652: Add error string to 'failed to resolve' exception |
00:52 | |
|
01:52
ilbot3 joined
05:43
dogbert17 joined
06:44
lizmat joined
08:05
domidumont joined
08:10
domidumont joined
08:44
dogbert17 joined
|
|||
| samcv | jnthn, I fixed the MVMROOT macro so it doesn't have that comma problem. turns out you were missing a semicolon | 09:36 | |
| well i added a semicolon to the MVMROOT macro which helps but didn't solve all the problems. and then adding parens around the { } fixes the rest. you need both changes to get it working | 09:38 | ||
| timotimo | i'm not going to be here for today, but it'd be a good idea to merge the branch "refuse_dangerous_inlines" before the release (it might want a little bit of attention before a merge) | 10:10 | |
| samcv | what's it do? | 10:12 | |
| i think i have almost fixed the concat issue | 10:13 | ||
| release blocker | |||
| yay i fixed it! | 10:14 | ||
| took me a long time to figure out wtf was going on. but it was increasing the repetitions since the last strand of a and first strand of b were the same | |||
| even though it created the renormalized section, so it ended up not using it and just increasing repetitions | 10:15 | ||
| which is why it was so damn hard to figure out | |||
| timotimo | well, spesh isn't written in a way that'd let it work with the GC | 10:21 | |
| i.e. if other threads want to do gc, spesh will finish up its stuff and let gc commence at a "safe spot" | 10:22 | ||
| i.e. where spesh isn't holding on to much memory | |||
| and when inlining, wval ops have to be fixed up because the inliner and inlinee may be from different compilation units, where each compilation unit could have a different target serialization context at the same index | |||
| the inlinee's CU will "disappear", so the object has to be "stolen", but if the object in the inlinee's CU's depended-on SC hasn't been deserialized yet, that'd cause an object to be allocated | 10:23 | ||
| and that can cause the GC to run, which currently is set up to panic moarvm | |||
| (so we shut down in an orderly fashion rather than corrupting memory and doing god-knows-what) | 10:24 | ||
| the "refuse dangerous inlines" branch simply checks if the object at that index is still to be deserialized and if it is it refuses to do the inline | |||
| that's far from optimal, but it's at least safe | |||
| timotimo AFK for long time (but maybe available on phone) | 10:25 | ||
| Geth | MoarVM: a8658127cd | (Samantha McVey)++ | src/strings/ops.c improvements to NFG_CHECK |
10:44 | |
| MoarVM: 5363a0742d | (Samantha McVey)++ | src/strings/ops.c Fix concatenation bug when renormalization needed This bug has only ever been seen in a precompiled module, because to trigger, you have to be concatenating string a and string b, where string a is a strand and string b are strands and the last strand of string a returns true with MVM_string_equal_at(). With this change, it checks if it can increase the repetitions if concat_stable == 1 |
|||
| nine | Why oh why did I think that writing assembly was a good way to relax on a weekend? | 11:03 | |
| samcv | night everyone | 11:05 | |
| jnthn | 'night, samcv++ | 11:06 | |
| nine | Actually assembly is innocent. It's really just the difficulty of finding non-confusing names for the operands of setcodeobj ;) | 11:07 | |
| What the? So I added support for setcodeobj and captureinnerlex to the JIT as they were blocking JITing of Block.clone. But now according to the profiler this method is not even speshed anymore? How can that happen? | 11:17 | ||
| Geth | MoarVM: f22f8d7c8b | (Stefan Seifert)++ | 2 files JIT setcodeobj and captureinnerlex These blocked JITing of Block.clone. JITing them may help closures. |
11:38 | |
| nine | Apparently there are now fewer speshed and more JITed frames (to be expected) but also fewer inlined frames. | 11:39 | |
| AlexDaniel | is there any policy on pushing non-bugfix changes like this ? before the release? | 11:49 | |
| because toaster will run with versions specified in NQP/MOAR_REVISION, any commits after that have unknown status | 11:51 | ||
| jnthn | "If you screw it up, you do the point release" :P | ||
| lunch; bbiab | |||
| robertle | annika | 12:14 | |
| annika]z~TUJM7gfUH\\HJL\\\\\\\\\\\\\\ | 12:16 | ||
| ~5~]~~ | |||
| sorry, children got to my keyboard | 12:17 | ||
|
12:42
domidumont joined
13:51
domidumont joined
|
|||
| [Coke] | AlexDaniel: at some point, we probably want to switch to a develop/master branching strat, where develop is auto-promoted to master once sufficient testing occurs. | 16:59 | |
| AlexDaniel | yea, that'd be cool | ||
|
17:07
domidumont joined
18:16
zakharyas joined
18:21
zakharyas joined
|
|||
| jnthn | AlexDaniel: At the point we want a release cut yet, or not? | 18:53 | |
| AlexDaniel | jnthn: there's only one thing that's worrying me: irclog.perlgeek.de/perl6-dev/2017-...i_15042503 | 18:54 | |
| jnthn: but I have no idea what to do with it | |||
| jnthn | Me either, from that information | 18:55 | |
| If it's some module and it's (mis)using NativeCall it might not even by on us to fix, for example... | 18:56 | ||
| And the error could be from anything | |||
| AlexDaniel | I'm currently testing modules that are marked as “burnt” | 18:58 | |
| would be nice if I catch at least one segfault there | |||
| if not, I guess we'll just ship as is, because what you gonna do? | |||
| most of them failed for completely different reason though | 18:59 | ||
| jnthn | I don't have time to do anything much this weekend besides cut the release | ||
| AlexDaniel | jnthn: alright. Then go for it | 19:00 | |
| jnthn | And s/this weekend/for the rest of August/ is almost as true...vacation from Tuesday :) | ||
| Alright | |||
| $ git describe | 19:01 | ||
| 2017.07-467-gf22f8d7 | |||
| ...this is gonna be some ChangeLog work :P | |||
| AlexDaniel | oh, there's a changelog for that too… :) | ||
| MasterDuke | jnthn: do you think you'll have time to look at github.com/MoarVM/MoarVM/pull/631 and github.com/MoarVM/MoarVM/pull/632 sometime ? or is there someone else who knows the relevant code well enough to review them? | 19:13 | |
| Geth | MoarVM: 28cb2cc6ae | (Jonathan Worthington)++ | docs/ChangeLog ChangeLog for 2017.08. |
19:44 | |
| jnthn | Review of that welcome | ||
| MasterDuke | "thershold" | 19:45 | |
| a couple one-line entries have a period, but most don't | 19:47 | ||
| AlexDaniel | jnthn: ping me before making the final step | ||
| Geth | MoarVM: 8f4f338cab | (Jonathan Worthington)++ | docs/ChangeLog ChangeLog fixes; MasterDuke++. |
19:50 | |
| AlexDaniel | errr | 19:51 | |
| c: 1599873d1420f gist.githubusercontent.com/AlexDan...tfile1.txt | |||
| committable6 | AlexDaniel, Successfully fetched the code from the provided URL. | ||
| AlexDaniel, gist.github.com/d8330f3c8a56f1aac0...bee4cac220 | |||
| AlexDaniel | c: 1599873d1420f^ gist.githubusercontent.com/AlexDan...tfile1.txt | ||
| committable6 | AlexDaniel, Successfully fetched the code from the provided URL. | 19:52 | |
| AlexDaniel, gist.github.com/066ba7691fc60e83ec...7b2ccdb87e | |||
| AlexDaniel | Geth: ver 1599873d1420f | ||
| Geth: ver github.com/rakudo/rakudo/commit/15...866cbfd87e | |||
| Geth | AlexDaniel, version bump brought in these changes: github.com/perl6/nqp/compare/2017....4-gf34562f | ||
| AlexDaniel | alright, not a moarvm thing it seems! | 19:53 | |
| back to #perl6-dev then :) | |||
| Geth | MoarVM: d1fc3a6bb1 | (Jonathan Worthington)++ | VERSION Bump VERSION. |
||
| jnthn | AlexDaniel: I've largely got things prepared. | 19:54 | |
| AlexDaniel | jnthn: I think there's an issue with github.com/perl6/nqp/commit/f34562...27706b9b66 | ||
| jnthn: it shouldn't really matter to you | |||
| jnthn | Hmmm | ||
| Yeah, that's up in NQP | 19:55 | ||
| Also apparently the JVM build mighta gotten busted by the atomics stuff, but I thought I stuck #?if moar around the lot of it :S | |||
| That's also NQP level though | |||
| In theory just copy-pasting a constant | |||
| Oh, maybe I know why it wasn't enough to just #?if moar the stuff I did | |||
| Anyway, will cut the MoarVM releae in an hour-ish :) | 19:56 | ||
| afk until then | |||
| AlexDaniel | jnthn: “optimazation” | 20:19 | |
| Geth | MoarVM: a425903282 | (Jonathan Worthington)++ | docs/ChangeLog Fix typo; AlexDaniel++. |
21:18 | |
| jnthn | AlexDaniel: Doing release now | 21:26 | |
| AlexDaniel | jnthn: ? | 21:27 | |
| jnthn | www.moarvm.org/releases/MoarVM-2017.08.tar.gz | 21:31 | |
| And tagged | |||
|
21:32
geekosaur joined
|
|||
| AlexDaniel | great. jnthn++ | 21:32 | |
| jnthn | AlexDaniel++ for the rest of the release stuff :) | 21:34 | |
| nine | Now THAT is a long changelog | 22:09 | |
| jnthn | Ain't it just! | 22:12 | |
| timotimo | "Move annotations multiple basic blocks if needed" | 22:42 | |
| ? | |||
| oh, like, move it "across" multiple basic blocks? | |||
| JIT of the new atomic ops - wants to perhaps have an "all"? | 22:44 | ||
| well, the release is already done | 22:46 | ||
| a pretty good-looking release this time :) | |||
| i regret that i didn't manage to merge the refuse_dangerous_inline branch before release | 23:19 | ||
| especially since dogbert was able to trip it in the spec test suite | |||
| anyway, i'll try to go a sleep | |||