| ab5tract | Needed to poke the optimizer as well of course | 00:35 | |
|
01:30
librasteve_ left
08:44
sena_kun joined
|
|||
| Geth | rakudo/main: c972df6865 | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | src/core.c/Supply-coercers.rakumod Allow a value of 0 with :vent-at Hopefully fixing #5358 |
09:39 | |
|
09:57
sena_kun left
|
|||
| Geth | rakudo/main: a917cf0321 | (Elizabeth Mattijsen)++ | src/core.c/Rakudo/Internals.rakumod Don't use ANSI color in error messages always but also stop doing that if STDERR is not a TTY. Fixes #5867 |
10:05 | |
| rakudo/main: fa72e6be9c | (Elizabeth Mattijsen)++ | t/12-rakuast/xx-fixed-in-rakuast.rakutest Make sure test compiles without RakuAST |
10:07 | ||
| lizmat | meh #5687, not 5867 | 10:08 | |
|
10:39
librasteve_ joined
|
|||
| ab5tract | when it comes to patching optimizer, which do you reckon is preferrable? gist.github.com/ab5tract/676be98f9...cafa7bf207 | 14:31 | |
| my gut says the all-QAST option is better? | |||
| vrurg | ab5tract: The less code to execute you leave for the runtime – the better. Therefore all-QAST is worse, in my view. | 14:40 | |
| ab5tract | ok, fair | ||
| it wasn't clear to me if the over-head of doing it in the optimizer might be larger than the check in QAST | 14:42 | ||
| "it" meaning the check | |||
| vrurg | The optimizer is executed once, at compile time. While the code it produces might end up in a hot path of runtime. | 14:45 | |
| ab5tract | seems pretty obvious when you put it that way :) | 14:49 | |
| my defense is that I've gotten 4 or less hours of sleep for the last week 😅 | |||
| vrurg | It takes time to wrap one's head around this stuff anyway. But then it clicks at some point... | 14:58 | |
| ab5tract | vrurg: I guess the only potential issue is whether there could ever be a scenario where the lexical `$?LANGUAGE-REVISION` could be different than the revision of the compiler that is doing the optimizing? | 15:27 | |
| I think it's highly unlikely, but as demonstrated my brain isn't 100% reliable at the moment | 15:28 | ||
| vrurg | You just should rely on compiler's interface for this. | 15:33 | |
| But, anyway, I can think of a scenario where QAST passed into optimizer is created with different compiler version, but it involves so much of handwork that I barely can figure out why would someone ever need this. | 15:38 | ||
| ab5tract | got it, thanks! | 15:43 | |
| donaldh | timo: tbrowder: could someone set up a wordpress advent account for me? | 16:03 | |
|
16:11
rba joined
|
|||
| Geth | rakudo/fix-negated-junctions: 72a7ef2996 | ab5tract++ | 5 files Address long-standing discrepancy for != and METAOP_NEGATE This patch uses revision-gate to migrate v6e to a version of negation that preserves junctions while accomplishing the same underlying functionality as planting `not` in front of `$expr`. Previously: ... (17 more lines) |
16:11 | |
| rakudo: ab5tract++ created pull request #5690: Address long-standing discrepancy for != and METAOP_NEGATE |
|||
| lizmat | donaldh: what WP username do you have? or what email address should I use ? | 16:16 | |
| donaldh ? in that case, you should have gotten an invite | 16:17 | ||
| donaldh | lizmat: my WP username is donaldhunter | 16:20 | |
| lizmat | invite sent | 16:21 | |
| donaldh | ty lizmat++ | ||
| timo | huh, libuv has tty support that translates ansi escape sequences in what you output to the equivalent on windows? | 18:41 | |
| docs.libuv.org/en/v1.x/guide/utilities.html#tty | |||
| lizmat | oooooh! | 18:43 | |
| timo | less necessary now that Windows Terminal or whatever it's called is Out There, but nonetheless ... | 18:50 | |
| lizmat: is there another change needed to make the ==Sorry!== text also not spit out ansi escape sequences? | 20:07 | ||
|
21:33
sena_kun joined
22:14
sena_kun left
|
|||