github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
MasterDuke jnthn, samcv, et al.: seems like it might be a good idea for someone to mention perl 6 in the hackernews commentary about the swift string atricle i mentioned the other day news.ycombinator.com/item?id=18394640 00:20
timotimo: you might also find hacks.mozilla.org/2018/11/cross-la...peedscope/ interesting 00:23
timotimo cool 00:55
nine dogbert11: the next commit is "Working bootstrap files" so I don't think b8ecfa43 is gonna build 11:49
dogbert11 nine: true, it was an amateurish attempt by me where I tried older and older nqp commits trying to figure out when the build hang on 32 bit systems began 12:12
given that this is a reasonable way to debug this (doubtful), commit d373f655, i.e. 'Working bootstrap files' is the first commit which causes the nqp build to hang without any kind of error message 12:16
on 32 bit systems, it still works on 64 bit ones 12:17
timotimo i wonder if for some reason building bootstrap files on a 32bit system makes them not work on a 64bit system
that'd obviously be a very bad bug 12:18
dogbert11 perhaps my way of 'bisecting' is not the correct one :) 12:19
dogbert11 hmm, I lied (or I'm doing something totally bogus) because I can't build nqp commit d373f655 on my 64 bit rig either. I get 'MVMHash representation requires MVMString keys - at gen/moar/stage2/QAST.nqp:7014 (src/vm/moar/stage0/QAST.moarvm:add)' 12:31
timotimo: are you still around? 12:32
timotimo o/
dogbert11 cool, am I doing the correct thing when tryin to build nqp
timotimo i'm not entirely sure :( 12:33
dogbert11 1 - go to nqp dir 2 - checkout interesting commit 3 - perl Configure.pl --backends=moar --gen-moar 4 - make (here it fails)
timotimo --gen-moar will not rebuild moar if it is newer than what you have 12:34
you'll want to force the exact commit from the MOAR_REVISION to be checked out 12:35
dogbert11 yeah, to be honest I have also tried step 2.5 - rm install/bin/moar 12:37
timotimo i don't think that's enough 12:40
i think it might look at the git repo to see what commit is there?
hum
no, that doesn't sound like a thing that we'd do
dogbert11 it's a mystery :) 12:51
nine dogbert11: unfortunately the changes I did are barely bisectable at best. The changes in the nqp an MoarVM repos are too closely tied and it can be quite a challenge to get that stuff to build. 12:56
I think plain old debugging will help best here. 12:57
dogbert11 nine: darn; i was hoping that I might be able to save you some time by doing a little prep work :( 12:59
nine I appreciate that very much :)
timotimo moarperf now has a search function for the callgraph, at least in the back-end. 19:00
brrt \o 19:58
timotimo the front-end has the search function now, too. 22:48
lizmat timotimo++ 22:49
timotimo there'll be a blog post
lizmat ++timotimo :-)
timotimo o/
\o
timotimo whoops, a profile with multiple million call graph nodes will not have its call graph data made searchable in just a few seconds ... 23:50