github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
japhb timotimo: Yeah, I think so. 00:05
Geth MoarVM: 72befa1647 | (Timo Paulssen)++ | src/jit/core_templates.expr
The exprjit of sp_bind_o caused wrongness

even though the sp_bind_o looked correct to many eyes. we expect the problem is something that used to not go through the exprjit because sp_bind_o caused the exprjit to bail out earlier is now active		 00:10
timotimo damnit, the other stuff got merged into moar master directly after the bump 00:19
so i'll have to also revert that and re-revert it back
Geth MoarVM: b2dc3f3ba2 | (Timo Paulssen)++ | 19 files
Revert "Merge pull request #1139 from MasterDuke17/add_context_to_error_messages"

This reverts commit bfe499dd6b891bc5101543337a7082ad356d0b16, reversing changes made to 2ac137c3d879e7e77c37522fcd7661fd9e4f6203.		 00:23
travis-ci MoarVM build passed. Timo Paulssen 'The exprjit of sp_bind_o caused wrongness 00:28
travis-ci.org/MoarVM/MoarVM/builds/585351219 github.com/MoarVM/MoarVM/compare/b...befa1647f7
travis-ci MoarVM build passed. Timo Paulssen 'Revert "Merge pull request #1139 from MasterDuke17/add_context_to_error_messages" 00:42
travis-ci.org/MoarVM/MoarVM/builds/585353748 github.com/MoarVM/MoarVM/compare/7...dc3f3ba2f0
timotimo really gotta look into this "coz compiler" thing 00:43
especially seeing if we can steal the way it works and put it into moarvm 00:44
nwc10 good *, #moarvm 06:44
brrt good * nwc10 06:54
timotimo uh oh, looks like more than just sp_bind_o is wrong 10:48
i wonder if i should revert all of them for now and put them into a branch 10:49
nwc10 suggests
1) step away from the keyboard
2) make coffee (or something)
3) ponder this
and after a delay, see if you have the same answer in your head
timotimo i've just been away from the keyboard for a whole bunch of hours :P
nwc10 aha. pre-delayed 10:50
as to the actual question, I don't know
timotimo i regularly stock up on away-from-keyboard time
nwc10 I'm not sure what I'm going to do for away-from-keyboard time once I've finished flattening the garden
timotimo terminate called after throwing an instance of 'dwarf::format_error' 11:20
what(): DW_FORM_sec_offset not expected for attribute (DW_AT)0x2119
:|
nine timotimo: moving them off into a branch would indeed be nice for others who work on the VM. Especially since they cause mysterious and hard to reproduce issues. 11:53
Geth MoarVM: 986e05beab | (Timo Paulssen)++ | 2 files
Revert "exprjit sp_p6oget_bi"

This reverts commit 0674a1e6940e8c7945dd6e5620ad3edc28414f43.		 11:56
MoarVM: a0ff25c746 | (Timo Paulssen)++ | src/jit/core_templates.expr
Revert "exprjit: templates for sp_get_i* and sp_get_o"

This reverts commit acfb5949967065694a058a31c159f99bfb730469.
nine Oh, I think I misdiagnosed the spesh plugin race condition issue. I thought that resolve_using_guards was using a plugin_state that was already freed by a concurrent add_resolution_to_guard_set. But this already uses MVM_fixed_size_free_at_safepoint. 12:39
I bet what actually happens is that this old plugin_state is no longer anchored somewhere where the GC would find it, so the guards it contains can become outdated. 12:41
Usually it would hang off the MVMStaticFrameSpesh but add_resolution_to_guard_set replaces it with the new version. 12:42
The bad news is that this is not just a case of adding an MVMROOT and be done with it. The good news is that this is still in the realm of problems that I know how to solve :) 12:44
Guest15407 timotimo: your templates are probably good, the problem lies elsewhere 12:49
timotimo yeah, but making them available causes trouble
Guest15407 true true :)
nine: what do you make of os_handle here? github.com/MoarVM/MoarVM/blob/mast...ops.c#L496 12:50
nine Guest15407: looks suspicious. Unless someone else keeps req->data up to date 13:19
Guest15407 nine: I have seen the mutex (line 500) suddenly become 0x0, see gist.github.com/dogbert17/bd5a41e0...d8a1536915 14:05
lizmat and yet another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2019/09/16/...essinging/ 15:41
MasterDuke timotimo: why did my PR merge have to be reverted? couldn't just the template commits have been reverted and then a new bump? 16:46
timotimo then the bump would have had the merge in it, too 16:47
MasterDuke also to everybody, my bad on asking for the bump in the first place
timotimo but the merge was specifically delayed after the bump
MasterDuke oh, ha. right
heh, forgot about that. timotimo++
nine It occured to me that I totally wasted those GC fixes I did. I could have kept a pile of them, just pushing one per week and ensure my mention in the weekly for almost a year in advance! Then I could have become known in the community as the possibly corrupt memory guy...or....wait...nah, better not :D 18:04
MasterDuke so many different options for hyphens to change meaning there: possibly-corrupt memory guy, possibly corrupt-memory guy, possibly corrupt memory-guy, possibly corrupt-memory-guy, possibly-corrupt-memory guy, possibly-corrupt memory-guy, possibly-corrupt-memory-guy 18:08
nine I wonder...how terrible would this patch be? gist.github.com/niner/19c93f7b96ae...7986d10e76 19:20
brrt \o 20:42
timotimo nine: did you mean to root temp_root_ptr? 21:16

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!