github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
02:37 lucasb left 03:33 MasterDuke left 06:19 domidumont joined
nwc10 good *, #moarvm 06:42
nine Good Friday! (Fridays are a thing on IRC, aren't they?) 07:03
nwc10 I believe so. But they seem to come and go. 07:04
07:12 patrickb joined
nwc10 patrickb: so we go deeper down the rabbit hole... 07:16
patrickb good * (morning over here) :-)
nwc10 The C library pow() (at least for glibc) is accurate for all powers of 10 07:17
but attempting to build those powers of 10 with floating point multiplication by the algorithm used in the pow_i OP isn't.
sigh
and then I found
r: say 1e23
camelia Error while reading '/home/camelia/p6eval-token': Permission denied at /home/camelia/rakudo-j-inst/bin/eval-client.pl line 10.
9.9999999999999992e+22
patrickb Should I try that on mingw?
nwc10 no, no need
because we (er, I) probably need to fix that one first 07:18
the problem seems to be (have not tested this with a debugger) that 1e23 (and one other) end up as binary that grissu3 can't format. So it falls back to sprintf 07:19
and we have sprintf with %17g
and that's, um, not correct here
but likely %15g isn't correct for other values.
the soution seems to be to move (it all) to Ryū 07:20
which has a suitably licensed implentation written in C 07:21
and then tackle the previous step. Worst case is 309 8-byte values in a lookup table.
But that's not for today. 07:22
I have no coffee. What could possibly go wrong? :-)
patrickb oh. That sounds like a fair share of pain. I recall how much effort and reading up zoffix once did to arrive at the grissu3 solution we currently have... 07:23
nwc10: Thank you for championing this.
nwc10 It's getting worrying - I can look at code that has if (e == 0 and m ==0) { ... } and know what that means just as-is 07:25
patrickb ao
sorry keyboard switched configs 07:26
exponent =0 and mantissa = 0 -> it's a zero float?
nine patrickb: that would e too obvious :) 07:28
patrickb (typo pun :-)
nine ah I wish that was on purpose :) 07:29
nwc10 yes. 07:35
08:11 zakharyas joined 08:53 Geth joined
jnthn o/ 09:51
nwc10 \o 10:02
10:47 linkable6 left 10:49 linkable6 joined 11:23 zakharyas left 11:33 hankache joined 11:38 frost-lab joined 11:40 hankache left 11:51 frost-lab left 12:52 Kaeipi joined 12:54 Kaeipi left 12:55 Kaiepi left 13:35 sena_kun left 13:37 sena_kun joined 14:08 zakharyas joined 14:09 sena_kun left 14:11 sena_kun joined 15:00 cog left, cog joined 16:12 cog left 16:13 cog joined 16:25 domidumont left 18:32 MasterDuke joined
MasterDuke nine: any reason you didn't PR github.com/MoarVM/MoarVM/commit/74505ce54d ? 19:15
19:18 MasterDuke left 19:19 MasterDuke joined 19:34 zakharyas left 19:35 zakharyas joined
sena_kun Somewhere between github.com/MoarVM/MoarVM/commit/ba...a5f0aaa47c and github.com/MoarVM/MoarVM/commit/94...518c590b92 the App::Uni module was broken, we need to investigate and fix it. 19:42
MasterDuke it's definitely a moarvm-caused problem, not all the unicode-related commits in rakudo between that timespan? 19:45
hm, but there are some string ops commits in there... 19:47
sena_kun MasterDuke, it bisects to the moarvm bump and I checked manually it's indeed the bump. 19:48
MasterDuke oh, but i just cloned it and the tests pass for me 19:49
hm, i wasn't at HEAD everywhere, let me try again... 19:51
yup, still pass 19:54
hm, that blin gist says it's version 1.0.2, but the git clone says 1.0.1... 19:56
oh, but its META6.json does say 1.0.2
[Coke]: does it fail locally for you? 19:58
also installs fine with zef 20:00
[Coke]: also, the repo description still says "Perl 6 - command-line utility" and the readme says unicode 12, but we're now using 13.1 20:08
japhb [Coke]: Any interest in PRs to merge in some of the functionality from gist.github.com/japhb/7b8a68da1b4b...c1269b951a ? 20:15
20:22 zakharyas left 20:54 patrickb left
nine MasterDuke: wait, I didn't? 20:55
MasterDuke: oh, I simply forgot after pushing the branch. But... there's more to come on that anyway. I found other places where we missed propagating the facts. Will commit and push those and then create the PR 20:57
MasterDuke cool