|
Parrot 4.8.0 "Spix's Macaw" | parrot.org/ | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Tuesday 19:30 UTC Set by moderator on 8 November 2012. |
|||
| kid51 loves to see whiteknight committing (again) | 00:09 | ||
| whiteknight | I finally had something that I felt like I really needed to do | 00:10 | |
| great motivation to get out of a slump | |||
| something is broken with pmc props | 00:16 | ||
| no, false alarm. a rebuild fixes it | 00:21 | ||
| dalek | rrot/whiteknight/io_vtable_lookup: c3edcbd | Whiteknight++ | / (6 files): Several fixes so the branch builds and passes most coretests. packfileconstanttable.t fails for some odd-looking reasons. I have not made test yet. |
00:24 | |
| cotto | Reini, any luck with the threads bug on Darwin? | 00:26 | |
| I really want to merge that branch, but I also don't want to cause problems for kid51's box. | 00:36 | ||
| Coke | is this the PPC box? | 00:41 | |
| (OS X PPC is probably out of support from apple at this point, given the # of intel only releases of the OS) | |||
| cotto | He said that it happens on x86_64. | 00:42 | |
|
01:16
Mike-PerlRecruiter_ joined
04:19
Reini_ joined
04:22
Reini_ joined
09:08
zby_home joined
10:29
moritz joined
|
|||
| dalek | kudo/nom: cecc9fa | moritz++ | src/Perl6/Actions.pm: replace QAST::VM with QAST::Op |
11:16 | |
|
13:13
PacoAir joined
13:17
Mike-PerlRecruiter_ joined
13:22
whiteknight joined
|
|||
| dalek | kudo/qast-sink-1: 5ad032d | moritz++ | src/Perl6/Actions.pm: first cut at sink context compiles the setting, but fails during startup |
13:36 | |
| whiteknight | good morning, #parrot | 13:43 | |
|
15:41
Reini_ joined
16:48
tuxit joined
17:15
Reini_ joined
|
|||
| dalek | kudo/qast-sink-1: 3f12074 | moritz++ | src/ (3 files): only sink when --sink command line option is provided |
18:16 | |
| Reini_ | cotto: Not yet. Still busy in another world | 18:27 | |
| whiteknight: There is more than just packfileconstanttable.t failing. See smolder.parrot.org/app/projects/rep...ails/32309 | 18:28 | ||
| cotto: Looks like I can now reproduce the 2 remaining threads errors | 18:39 | ||
| t/src/embed_2 fails in puts("Done") | 19:06 | ||
| Is puts not thread-safe on darwin? | |||
| fprintf(stdout also fails. Parrot_interp_allocate_interpreter is wrong without parent for threads | 19:17 | ||
| msg whiteknight Do you have access to darwin? #856 is your domain. puts fails in t/src/embed_2.c. Somehow Parrot_interp_allocate_interpreter without parent corrupts IO. | 19:19 | ||
| aloha | OK. I'll deliver the message. | ||
| Reini_ | msg whiteknight See github.com/parrot/parrot/issues/856 | 19:24 | |
| aloha | OK. I'll deliver the message. | ||
|
19:25
bouncy joined
19:26
contingencyplan joined
|
|||
| dalek | Heuristic branch merge: pushed 24 commits to parrot/threads by rurban | 19:35 | |
|
19:47
travis-ci joined
|
|||
| travis-ci | [travis-ci] parrot/parrot#632 (threads - 524efde : Reini Urban): The build was fixed. | 19:47 | |
| [travis-ci] Change view : github.com/parrot/parrot/compare/c...4efdecf459 | |||
| [travis-ci] Build details : travis-ci.org/parrot/parrot/builds/3254145 | |||
|
19:47
travis-ci left
|
|||
| dalek | rrot/threads: e9dea3c | rurban++ | t/src/e (2 files): fix [GH #856]: a 2nd interp on threaded parrot needs a parent On darwin IO fails when a second interp has no parent. Disable the failing tests for now. To fix this we would need to change the specs for Parrot_interp_new that either the interp are automatically connected, are subsequent NULL (empty parent) are disallowed. |
19:50 | |
| Reini_ | s/are subsequent/or subsequent/ | 19:51 | |
| threads pass now. we need to discuss this with our customers (rakudo, nap, mod_parrot, …) | 19:52 | ||
| tadzik | Reini_: did you try my testcase on nqp? | ||
| gist.github.com/3817314 | 19:54 | ||
| Reini_ | I think I have a better threads fix now. | 20:03 | |
| tadzik: not yet | |||
| I connect now all subsequent interp as kids of the first | 20:06 | ||
| emergency_interp is the new default parrent | 20:07 | ||
| cotto | ~~ | 20:12 | |
|
20:14
brrt joined
|
|||
| dalek | rrot/threads: 24f8548 | rurban++ | / (3 files): [GH #856] Clarify the Parrot_interp_new API for subsequent empty parents Fixed the last blocking threads bugs, by automatically connecting subsequent new interpreters to the very first interpreter. Revert the test changes in the previous commit. t/src/extend_19 still fails for mysterious reasons, manually it passes, so TODO it. |
20:18 | |
| Reini_ | The API for Parrot_interp_new already suggested that only the first can have an empty parent. I clarified that now. | 20:19 | |
| Looks like threads can now be merged :) | 20:20 | ||
| cotto | ! | 20:23 | |
| sorear | Again? | 20:24 | |
| tadzik | :D | ||
| Reini_ | I'll do some extensive tests now, also with rakudo | 20:26 | |
| cotto | same here | 20:27 | |
|
20:30
travis-ci joined
|
|||
| travis-ci | [travis-ci] parrot/parrot#634 (threads - 24f8548 : Reini Urban): The build was broken. | 20:30 | |
| [travis-ci] Change view : github.com/parrot/parrot/compare/e...f8548cb33c | |||
| [travis-ci] Build details : travis-ci.org/parrot/parrot/builds/3254594 | |||
|
20:30
travis-ci left
20:32
brrt joined
|
|||
| Reini_ | tadzik: your case still fails. bad | 20:42 | |
| cotto | tadzik, that test can't work until the test libraries are updated to be thread-aware. | 20:43 | |
| I'm planning on doing that as soon as I understand the branch well enough. | |||
| Reini_ | tadzik's case was just gist.github.com/3817314 | 20:44 | |
| no Test::More | |||
| cotto | oh | 20:45 | |
| read that too quickly | |||
| dalek | rrot/threads: 316916a | rurban++ | src/interp/api.c: [GH #856] init the first emergeny_interp parent properly, hard tabs Remove hard tabs. Init emergency_interp->parent_interpreter = NULL |
20:50 | |
| Reini_ | whiteknight was analyzing tadzik's case | 20:51 | |
| cotto | I can't build nqp against the threads branch. I'm verifying that it's the branch's fault. | 20:52 | |
| looks like nqp's fault | |||
| maybe the gcc that comes with ubuntu 12.10 is pickier than previous versions | 20:57 | ||
| adding -fpermissive to gcc | 20:58 | ||
| 's flags gets it working again | |||
| well, more working | 20:59 | ||
|
21:02
travis-ci joined
|
|||
| travis-ci | [travis-ci] parrot/parrot#635 (threads - 316916a : Reini Urban): The build was fixed. | 21:02 | |
| [travis-ci] Change view : github.com/parrot/parrot/compare/2...6916aa13c1 | |||
| [travis-ci] Build details : travis-ci.org/parrot/parrot/builds/3255045 | |||
|
21:02
travis-ci left
|
|||
| cotto | and apparently gcc 4.7.2 doesn't like "typeid" as a variable name | 21:02 | |
| tadzik++'s code still goes boom | 21:08 | ||
|
21:25
Reini_ joined
|
|||
| cotto | Adding -fpermissive to the list of gcc flags that Parrot uses makes nqp's casting problems go away. | 21:27 | |
| any reason not to do that? | |||
| dalek | kudo/qast-sink-1: 5574863 | moritz++ | src/core/Failure.pm: Failure.sink makes the exception blow up |
21:34 | |
| kudo/qast-sink-1: 3ee7808 | moritz++ | src/ (2 files): unbreak Hash/Block distinction |
|||
|
21:44
brrt1 joined
|
|||
| Reini_ | testing now with gcc-4.8 | 22:37 | |
| dalek | rrot: 6d516e4 | cotto++ | config/auto/warnings.pm: add -fpermissive so that nqp builds under gcc 4.7.2 |
22:42 | |
|
22:44
brrt1 left
22:50
whiteknight joined
|
|||
| cotto | hio whiteknight | 22:52 | |
| whiteknight | hey cotto | ||
| cotto | It's getting close to merge o'clock for the threads branch. | ||
| nqp passes its tests under the parrot threads branch with one change that I mentioned in #perl6 | 23:03 | ||
| whiteknight | oh really, what change was that? | 23:08 | |
| Also, I wasn't aware that NQP failed any tests in that branch, I just thought some of the lexpad stuff failed if the threads were actually used | 23:09 | ||
| but for single-threaded case I thought it worked the same as before | |||
| Also, I'm 75% of the way to some big improvements to the IO system (again). I'll send out an email soon with some details of what I'm doing and why | 23:12 | ||
| The first big change we have to talk about, if we want to buy what I'm selling, is to split FileHandle up into a new Pipe PMC | |||
| The ultimate goal is that we'll be able to treat any arbitrary PMC as a Handle type, without having to derive from or delegate to any existing handle type | 23:16 | ||
| cotto | It's no a parrot-specific change. | 23:18 | |
| gcc 4.7 doesn't like one of its variable names | |||
| *not | |||
| whiteknight | oh, okay. So with that gcc fix, nqp builds fine? | ||
| cotto | yup and passes its tests | ||
| whiteknight | Okay, good | ||
| cotto | rakudo also builds | 23:19 | |
| running spectest_regression now | |||
| whiteknight | I know that nqp lexpads don't work with threads yet. But I think we can still merge the branch and just tell the nqp folks that they just can't use them yet until we update nqp to be thread-safe | ||
|
23:21
Hunger joined
|
|||
| cotto | we're only at the stage of making sure that we don't break Rakudo when we merge | 23:22 | |
| whiteknight | yeah | 23:23 | |
| dalek | rrot/whiteknight/io_userhandle: 48f6e2e | Whiteknight++ | src/io/userhandle.c: Update userhandle.c to redirect most low-level IO vtable routine requests to named methods on the handle object. |
||
| rrot/whiteknight/io_userhandle: fe092f7 | Whiteknight++ | / (2 files): Fixes so the system builds |
|||
| rrot/whiteknight/io_vtable_lookup: 849172b | Whiteknight++ | src/pmc/handle.pmc: Remove read_buffer and write_buffer from Handle PMC, since they aren't explicit attributes |
23:30 | ||
| cotto | seeing some test failures: gist.github.com/4108166 | 23:42 | |
| checking if they happen against master too | 23:43 | ||
| be back in a few hours | |||
| whiteknight | That's in the threads branch? | 23:45 | |