<kares[m]>
hey, you guys have your oss idea licenses renewed?
sagax has joined #jruby
dhoc has joined #jruby
subbu|away is now known as subbu
<enebo[m]>
kares do not quit idea
<enebo[m]>
kares: but yeah hiro asked last wednesday and has not hard back yet
<enebo[m]>
OR change the date on your machine :)
<kares[m]>
okay no biggie
<enebo[m]>
It only seems to start mentioning it will stop working like 4 days before it happens
<enebo[m]>
Or at least that was when I noticed it
<kares[m]>
yy
<kares[m]>
not quiting not an option, for some reason project windows aren't always closing (a blank window persists) until I restart
<enebo[m]>
kares: manual date I think should work but that is no fun
<enebo[m]>
another option could be to sign up with evaluation copy
<enebo[m]>
which may require another email address
<enebo[m]>
I suspect covid-19 has affected their productivity
dhoc has quit [Quit: dhoc]
dhoc has joined #jruby
xardion has quit [Remote host closed the connection]
xardion has joined #jruby
shellac has quit [Ping timeout: 260 seconds]
<headius[m]>
I quit and it started up again fine
<headius[m]>
couple days ago
<enebo[m]>
headius: I half wonder if they disabled license checking somehow
<enebo[m]>
I was too afraid to see :)
<enebo[m]>
I did look for a covid page but did not see one
<headius[m]>
yeah I have no idea
<headius[m]>
ok, TODAY we'll get 2.6 merged
<headius[m]>
not sure where these last few were failures are coming from... never seen them up to now
<headius[m]>
were=weird
<enebo[m]>
I am going to fix or attempt to unbreak for loops with serialization on master since that regressed when we merged our IRScope removal branch
<enebo[m]>
The second thing broken is ir inliner has no working callsites any more
<enebo[m]>
but for for now :)
<headius[m]>
for loops with serialization?
<headius[m]>
what madness is this
<headius[m]>
oh you mean serializing IR of for loops
<enebo[m]>
lvars read in from serialization end up with improper depth
<headius[m]>
I was imagining some unholy combination of for loops and Marshal
<headius[m]>
ok
<headius[m]>
I'm going to actually take a deeper look at these failures because they suddnely showed up in CI
<headius[m]>
some of them I can't repro locally, so not sure about that
<enebo[m]>
Ultimately I just want to read in staticscope and be done and non-zero depths will calculate from that
<headius[m]>
enebo: wow I found one that only fails in full interpreter
<enebo[m]>
nice
<headius[m]>
that lambda thing seemed to be that the optimized prepareArgs logic was splatting the incoming args array before arity checking, and the non-optimized path didn't do that at all
<headius[m]>
we need to figure out all these different prepareArgs variants and clean this up
<headius[m]>
there's like five different ways to prepare args for a block
<headius[m]>
I deleted the splatting and we'll see how it does in CI
<headius[m]>
and it's exactly two because the second call re-gets context which is now full
<enebo[m]>
yep
<headius[m]>
aha of course
<headius[m]>
it does promoteToFullBuild inside the call path for simple
<headius[m]>
so it can't go back
<enebo[m]>
Ok so threshold=0 is stll just moving through startup on acquire but then getting replaced from jit
<headius[m]>
this may be specific to blocks
<headius[m]>
yeah
<headius[m]>
so it does run synchronously but by the time we submit for JIT we've already committed to startup call path
<headius[m]>
this explains a lot of other rando failures too
<headius[m]>
like fstring idempotence
<headius[m]>
now about fixing it
<enebo[m]>
we could pobably make clone clone value but that is very worrisome to me :)
<headius[m]>
yeah worries me a little too
subbu is now known as subbu|away
<enebo[m]>
I mean this would be operands keep their values
<headius[m]>
right
<enebo[m]>
and that would be very weird for let's say a temp var
<headius[m]>
so we only create or lookup or cache once
<headius[m]>
but we can't apply that to bytecode jit in any case
<headius[m]>
I am thinking of tagging for now
<headius[m]>
I mean I could do some work to move the block jit checks earlier in dispatch but it's a bunch of code for a very specific and rare and pretty much benign case