01:09 kvw_5_ joined 01:12 kvw_5 left 01:51 frost-lab joined 02:16 lucasb left 07:08 sena_kun joined 09:50 ZzZombo joined 10:07 LizBot left 10:08 ZzZombo left 11:28 Xliff joined
Xliff \o 11:28
nwc10 o/
lizmat afk for a few hours& 11:40
Geth nqp/sprintf-pow10-lookup: dcfb06b3b7 | (Nicholas Clark)++ | tools/generate-sprintf-pow10-table.raku
A tool to generate a lookup table to accurately calculate pow_n(10.0, $exp).

We need this for floating point formatting in sprintf, and nqp::pow_n is not always accurate.
11:57
nqp/sprintf-pow10-lookup: 61c8759b1c | (Nicholas Clark)++ | src/HLL/sprintf.nqp
Accurately calculate pow_n(10.0, $exp) for stringify-to-precision2.

nqp::pow_n() is not always accurate; nor is looping and multiplying by 10 or any predictable pattern of powers of 10.
nqp/sprintf-pow10-lookup: 444a683339 | (Nicholas Clark)++ | src/HLL/sprintf.nqp
Convert `scientific` and `shortest` to use `pow10` and an `int` for `$exp`.

Previously both had typed `$exp` as `num`, even though it only holds integers.
nqp: nwc10++ created pull request #716:
Accurately calculate pow_n(10.0, $exp) for floating point sprintf formatting
12:03
Kaiepi o/ 13:12
i have some legal concerns about the work i've been doing on network addresses and dns resolvers 13:13
i use modified snippets of openbsd source code in a few spots. i know there are comments i need to worry about in doing so, but is this permissible, and if so, is there anything else i need to worry about? 13:15
sena_kun Kaiepi, you might ping codesections in #raku 13:18
MasterDuke my advice would be to purchase a new name and identity, move to someplace like Macau, and preemptively put out a hit on Theo de Raadt. but IANAL...
nine makes a mental note not to ask MasterDuke for legal advice... because he ain't a lawyer 14:00
14:02 frost-lab left
Kaiepi sena_kun, might want to ping? 14:10
sena_kun Kaiepi, yeah, sorry, "might want". 14:11
Kaiepi ok, thanks 14:12
14:15 Kaiepi left, Kaiepi joined 14:18 b2gills left, b2gills joined
MasterDuke patrickbkr[m]: what's the reason for running prove with `-j0` in the nqp test stage? the rakudo test doesn't specify any `-j` option 14:31
any reason not to use -j2 for `make` and `prove`? 14:33
14:38 Kaiepi left, Kaiepi joined 14:43 Kaiepi left 14:44 Kaiepi joined
MasterDuke anybody here on windows ever see this when running rakudo's tests? dev.azure.com/MoarVM/MoarVM/_build...amp;l=1900 15:24
the error is: Failed to rename 'D:\a\1\rakudo\t\04-nativecall\.precomp\3350F24AD6CA381ED1FFF13D57FAE45D527E2FB4\64\640AB2BAE07BEDC4C163F679A746F7AB7FB5D1FA.repo-id.tmp' to 'D:\a\1\rakudo\t\04-nativecall\.precomp\3350F24AD6CA381ED1FFF13D57FAE45D527E2FB4\64\640AB2BAE07BEDC4C163F679A746F7AB7FB5D1FA.repo-id': Failed to rename file: operation not 15:33
permitted
at D:\a\1\rakudo/t/04-nativecall/02-simple-args.t:6
Actually thrown at:
  in any statement_control at D:\a\1\install-moved\bin\..\share\perl6\lib/Perl6/Grammar.moarvm line 1
this is after i added `-j2` to the prove command 15:37
ugexe then its probably because one process is holding a file open while another is e.g. trying to close it 15:38
er, or renaming a file that another process has a handle open on
15:39 softmoth_ joined
MasterDuke yeah, is `-jn` just not safe on windows? 15:39
patrickbkr[m] MasterDuke: I have only ever used -j for the MoarVM build on Windows. 15:41
-j0 was probably faithfully copied from some older CI system I looked at when creating the Azure config... 15:42
MasterDuke so you haven't tried it with prove?
github.com/MoarVM/MoarVM/pull/1475...2269785e60 looks like it took 10m off the total runtime, so i'd like to keep as much of it as possible 15:43
ugexe well technically it seems that multi process on windows where each process uses some of the same modules isn't safe
MasterDuke huh, how has that not cropped up before? 15:44
ugexe it pops up fairly often. nine has fixed it a couple of times, but apparently some gremlins still lurk 15:45
windows file system time granularity and how lock file work on windows are the two things (i think) that make the whole thing work poorly on windows
well, not poorly. less than perfectly 15:46
raydiak windows antivirus can cause this according to github.com/valery-barysok/session-...671281388. try disabling it? the comment at the end of that same discussion suggests a permanent fix, if that's the issue 15:50
MasterDuke hm, leave the PR as is in the hope it forces someone (probably nine or you) to fix? or remove the `-j2` from prove on windows (even though that's where the biggest speedup was)?
raydiak: ooh, interesting... 15:51
ugexe personally (and im probably in the minority) i like to run my CI tests with just one job so its as reproducable as possible 15:52
well the way windows antivirus would cause the issue is the same way two raku processes trying to precompile would cause it
windows antivirus is just another thing that might have the handle open
MasterDuke heh, i guess both would be good. reproducibility with the single job pipeline, more chaos with the multiple job pipeline 15:54
[Tux] Rakudo v2021.03-183-g600bb6b06 (v6.d) on MoarVM 2021.03-50-g948128995
csv-ip5xs0.854 - 0.871
csv-ip5xs-208.208 - 8.905
csv-parser25.858 - 26.614
csv-test-xs-200.382 - 0.400
test7.308 - 8.028
test-t1.993 - 2.275
test-t --race0.893 - 0.948
test-t-2033.009 - 35.475
test-t-20 --race10.004 - 10.522
15:57
ugexe yeah running a e.g. -j12 on the entire suite could be considered a (separate) test in its own right (basically a naive stress test) 16:00
MasterDuke that's how i always run it locally
Geth ¦ rakudo: lizmat self-assigned SlippyIterator.is-deterministic() fails on fresh instance github.com/rakudo/rakudo/issues/4335 17:58
17:59 sena_kun left 18:00 sena_kun joined 19:13 MasterDuke left
Geth nqp/sprintf-pow10-lookup: 895c96d7fd | (Nicholas Clark)++ | tools/generate-sprintf-pow10-table.raku
A tool to generate a lookup table to accurately calculate pow_n(10.0, $exp).

We need this for floating point formatting in sprintf, and nqp::pow_n is not always accurate.
19:25
nqp/sprintf-pow10-lookup: f12b7e7156 | (Nicholas Clark)++ | src/HLL/sprintf.nqp
Accurately calculate pow_n(10.0, $exp) for stringify-to-precision2.

nqp::pow_n() is not always accurate; nor is looping and multiplying by 10 or any predictable pattern of powers of 10.
nqp/sprintf-pow10-lookup: 9a766ac827 | (Nicholas Clark)++ | src/HLL/sprintf.nqp
Convert `scientific` and `shortest` to use `pow10` and an `int` for `$exp`.

Previously both had typed `$exp` as `num`, even though it only holds integers.
19:27 softmoth_ left
[Coke] apparently needs to do more than skim those commits to understand them :) 19:36
nwc10 mine? 19:38
apologise for using rude words, but: 19:40
perl -e 'printf "%s: %a\n", $_, $_ for map { 10 ** $_ } 0 .. 30'
and there is no pattern in the bit patterns in the mantissas of the powers of 10s
(and also ** is not trustworthy everywhere - that's the zeroeth problem) 19:41
enough to mean that nqp::pow_n(10.0, $exp) isn't quite correct everywhere for all integer values of $exp
and this messes up some sprintf formats
I don't propose to add %a to the current sprintf implementation. 19:42
lizmat %a would be ?
nwc10 but once lizmat's new one has landed, I'll try to remember to figure it out
floating point in hex notation
lizmat ah
nwc10 turns out that it's useful for debugging
(And I guess also efficient round tripping to strings) 19:43
no normal program needs it
19:50 lucasb joined
Geth rakudo: 1c4577aed1 | (Elizabeth Mattijsen)++ | src/core.c/Rakudo/SlippyIterator.pm6
Fix SlippyIterator.is-deterministic

Make it always return False. As the class consuming the role can have multiple slippyiterators during its lifetime, as soon as one returns False, the whole of the iterator should be considered non-deterministic.
But that would require a time-machine, which we don't have. So do the ... (5 more lines)
20:04
20:15 MasterDuke joined 20:31 MasterDuke left 20:32 MasterDuke joined
sjn heya; I'm trying to remember, but wasn't there at some point an irc channel dedicated to documentation issues? 20:42
lizmat before my time, maybe ? 20:48
sjn Nah; I think I'm misremembering.... 20:59
21:05 Ven_de_Thiel joined 21:29 softmoth_ joined 21:34 softmoth_ left 21:43 Ven_de_Thiel left
Geth rakudo: c64dc57337 | (Elizabeth Mattijsen)++ | src/core.c/Str.pm6
Revert infix:<~> subclassability

After discussion with jnthn, it felt too early to define ways in which concatenation of one or more subclasses of Str should be handled. So for now, concatenating subclasses of Str will produce a Str, regardless of any subclasses of Str involved.
21:52
21:54 dogbert11 joined 21:58 dogbert17 left 22:02 dogbert17 joined 22:07 dogbert11 left 22:34 dogbert11 joined 22:37 dogbert17 left