| Geth | roast/6.c-errata: 80674ec44d | (Daniel Green)++ | S02-types/int-uint.t Porting the fix in aa42c8340b to 6.c-errata The quick overview is that the tests for (over|under)flow and assigning literals more than max or less than min were incorrect. |
00:15 | |
| IOninja | m: (2.5..0).int-bounds.say | 13:57 | |
| camelia | Cannot determine integer bounds in block <unit> at <tmp> line 1 Actually thrown at: in block <unit> at <tmp> line 1 |
||
| IOninja | c: HEAD~10 (2.5..0).int-bounds.say | ||
| committable6 | IOninja, ¦HEAD~10: «Cannot determine integer bounds in block <unit> at /tmp/bpAPdfnMyE line 1Actually thrown at: in block <unit> at /tmp/bpAPdfnMyE line 1 «exit code = 1»» | ||
| IOninja | reallly | ||
| star: (2.5..0).int-bounds.say | 13:58 | ||
| camelia | Cannot determine integer bounds in block <unit> at <tmp> line 1 Actually thrown at: in block <unit> at <tmp> line 1 |
||
| IOninja | Hm. Weird. I thought I broke this last night but looks like it was already broken | ||
| or not | |||
| star: (-2.5..0).int-bounds.say | 13:59 | ||
| camelia | Cannot determine integer bounds in block <unit> at <tmp> line 1 Actually thrown at: in block <unit> at <tmp> line 1 |
||
| IOninja shrugs | |||
| Will take a closer look when my Internet goes back up -_- | |||
| dammit.... called my ISP. Thry said there's no outage and it's likely some problem with the cable :( | 14:33 | ||
| The earliest they can come and fix this is Monday :'( | |||
| [Coke] | we have a few places in town with free wifi. I might be camped out there over the weekend. | 14:46 | |
| IOninja | Many free wifi places here too. imma just gonna tether to my for release and the rest of the time... be productive, I guess :) | 15:37 | |
| *to my phone | |||
| The Gods of Fiber Optic have answered my payers. My Internet's back up without any intervention by the ISP's tech :) | 16:51 | ||
| geekosaur | "payers", Freudian much? | 16:52 | |
| IOninja | :) | ||
| bartolin | we have multiple test files in roast where we run one test in a for loop for different input. one example is this: github.com/perl6/roast/blob/aa42c8...ice.t#L319 | 19:23 | |
| now it sometimes happens that the test passes for most iterations but failes for a few | |||
| in the example the test fails on the jvm backend for the last element of @test | 19:24 | ||
| my question is: what's the best way to fudge in such a case? | |||
| in the past I tried to pull out the problematic parts in order to be able to fudge them, like in roast commit 7e1ef3ee08 | 19:26 | ||
| but that's not something I'd like to do regularly. on the other hand I'd like to avoid spectest failures caused by such known problems | 19:27 | ||
| Geth | roast: c4f712d9de | usev6++ | S32-io/seek.t [JVM] Fudge failing tests It looks like seeking works ok, but there is a problem with LEAVE on rakudo-j. I had to put the subtests in extra blocks in order to fudge them properly. |
19:42 | |
| gfldex | m: multi sub foo(Failure){}; multi sub foo($_){.say}; dir('.', :recursive, :f, :!d)».&foo | 20:01 | |
| camelia | ".cpanm".IO ".local".IO ".npm".IO ".perl6".IO ".perlbrew".IO ".rcc".IO ".ssh".IO "Perlito".IO "evalbot".IO "log".IO "nqp-js".IO "p1".IO "p2".IO "perl5".IO "std".IO ".bash_history".IO ".bashrc".IO "mbox".IO ".lesshst".I… |
||
| IOninja | bartolin: what about fudging a bit more than necessary; all tests in the group or something? | ||
| gfldex | IOninja: I found myself wanting to do this ^^^ | ||
| IOninja | r: dd [ [|'I ❤️ Perl 6'.comb], 10, 6 ] | 20:02 | |
| camelia | [["I", " ", "❤️", " ", "P", "e", "r", "l", " ", "6"], 10, 6] | ||
| [["I", " ", "❤", "️", " ", "P", "e", "r", "l", " ", "6"], 10, 6] | |||
| IOninja | uh-oh | ||
| bartolin | IOninja: well, but then I'd get some passed TODO tests -- that's annoying as well | ||
| IOninja | m: dd 'I Perl 6'.comb | 20:03 | |
| camelia | ("I", " ", " ", "P", "e", "r", "l", " ", "6").Seq | ||
| bartolin | of course the could be skipped, but that feels a bit like overkill | ||
| IOninja | ugh my terminal is playing tricks on me | ||
| bartolin | *they | ||
| IOninja | bartolin: not sure what to suggest. I'm probably the author of that test and I use these loops so to not dupe the test and its description all the time. | 20:04 | |
| So if letting the test fail is bad and fudging too many is bad, that's the only alternative I see and rather not :( | 20:05 | ||
| That example is small, but many examples have dozens of test cases in an array all tested with a single line of a test, instead of dozens of test lines | |||
| bartolin: actually, in that particular case, looks like the failure is due to .comb on JVM being busted with unicode chars | 20:06 | ||
| timotimo | yo uknow you can "emit" code for a given backend with the fudger? | ||
| bartolin | timotimo: yes, but I don't see how that would help in this case. I'prefer to run those tests | 20:07 | |
| Geth | roast: 600028af7f | (Zoffix Znet)++ | S32-array/splice.t Toss Unicode char that's problematic on JVM Has no relation to the test itself. |
20:08 | |
| timotimo | @fudges = False xx $num-of-tests; #?jvm emit @fudges[10..*] = True | 20:09 | |
| bartolin | IOninja: thanks, that was another alternative to fudging | 20:10 | |
| timotimo: I have to think about that a bit :-) | 20:11 | ||
| bartolin was aware of emit, but didn't consider to use it that way | 20:12 | ||
| timotimo | :) | ||
| Geth | roast: 3e8da11435 | usev6++ | S32-list/squish.t [JVM] Run passing test that was masked by 'emit' |
20:30 | |
| bartolin | timotimo++ # making me look into emit | ||
| timotimo | yay | 20:31 | |
| Geth | rakudo/nom: 16ef21c162 | (Zoffix Znet)++ | src/core/Range.pm Fix regression in Range.int-bounds Yesterday's fix[^1] to the method assumed an .Int on endpoints will be False only if the conversion returns a Failure, but it's possible to construct a non-int Range with `0` as one of the end-points, whose .Int is False. Fix by adding additional type checks that .Int's are Int [1] github.com/rakudo/rakudo/commit/79...2acd18bec2 |
22:53 | |
| roast: c8767e2b93 | (Zoffix Znet)++ | S02-types/range.t Test Range.int-bounds works on non-int Range with 0 endpoint Rakudo fix: github.com/rakudo/rakudo/commit/16ef21c162 |
22:54 | ||
| roast/6.c-errata: 2edc8f282d | (Elizabeth Mattijsen)++ | 4 files Handle fallout of List.reverse -> Seq migration |
23:25 | ||
| IOninja | Well, now I feel a little justified in complaining about t/fudgeandrun being moved to roast instead of rakudo. It's not present in 6.c-errata... | 23:28 | |
| "fudgeandrun has moved to roast but does not seem to be there. Perhaps updating roast (t/spec) will help." | |||
| I hope all these people who are too lazy to run roast from a rakudo checkout are being helped by the change | 23:29 | ||
| </grumpy> | |||
| Geth | roast/6.c-errata: 7b0c5a77a3 | (Zoffix Znet)++ | fudgeandrun Add fudgeandrun script to 6.c-errata branch |
23:30 | |
| roast/6.c-errata: 4c16a7e405 | (Zoffix Znet)++ | S09-multidim/methods.t Fix plan count The cherry-pick[^1] of the List.reverse change[^2] missed that the plan was reduced by 2 tests. [1] github.com/perl6/roast/commit/2edc...dad9c98eeb [2] github.com/perl6/roast/commit/93bd...c6a78ff583 |
23:36 |