|
#parrot Parrot 4.11.0 "[All together - Happy Birthday Lovebird]" --with-threads | parrot.org/ | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Tuesday 19:30 UTC Set by moderator on 18 December 2012. |
|||
|
00:05
Psyche^ joined
01:14
Mike-PerlRecruiter_ joined
|
|||
| Reini | I'm now redesigning the build layout, so that executables work as installed: bin/ lib/ perl5/ No need for installables, relocatable and sane. | 01:33 | |
| dalek | rrot/new-layout: 37b34ac | rurban++ | / (3 files): [GH 883] Darwin uses now @executable_path/blib/lib/... DYLD_LIBRARY_PATH disabled, there is no need anymore. Add new INSTALLABLELIBPARROT target, which is only needed on darwin to set the final install_name. Enable tools/dev/mk_language_shell.pl test for Darwin, as this works now. What is missing is copying INSTALLABLELIBPARROT in the installer, and later provide relative @executable_path/../lib/libparrot-version.dylib to enable relocatable installations. |
04:29 | |
| rrot/new-layout: ca58bca | rurban++ | / (2 files): install blib/lib/installable_libparrot.4.10.0.dylib And some whitespace fixes in MANIFEST.generated |
|||
| rrot/new-layout: 6653428 | rurban++ | / (361 files): New directory layout lib/bin/perl5 binaries into bin/, libraries into lib/, perl libs into perl5/ rename frontend/parrot -> frontend/parrot_old, frontend/parrot2 -> frontend/parrot, This way installed binaries have the same relaive paths to its libraries, which eases installation and testing a lot. binaries get relocatable on platforms which support relative loader paths, and can be copied as is. This is also the normal way a typical project layout is organized. |
|||
|
06:44
MikeFair_ joined
08:28
Reini joined
09:03
Reini joined
10:05
Liz_ joined,
woolfy joined
10:08
Reini joined
11:09
bouncy joined
11:15
Reini joined
12:05
klaus joined
|
|||
| klaus | Configure.pl says that I have no pcre, GMP, OpenGL. How important is that / what part of parrot does that use? | 12:08 | |
|
12:17
Reini joined
|
|||
| klaus | in addition to the above Configure.pl reports libffi as lack pkg.config. what are the consequences of that? | 12:28 | |
|
13:14
Mike-PerlRecruiter_ joined
13:21
Reini joined
13:42
benabik joined,
Reini joined
|
|||
| benabik | ~~ | 13:42 | |
| rurban | klaus: GMP is needed for the bigint,bignum pmcs, the other are just optional. | 13:59 | |
| Nothing to worry about | |||
| klaus | thanks rurban | 14:00 | |
|
14:10
Reini joined
|
|||
| klaus | although 'make test' prints 'All tests successful' it produces a segfault according to the output of dmesg (Ubuntu 12.10 x86-64) | 14:16 | |
|
14:18
PacoAir joined
|
|||
| rurban | at which test? | 14:19 | |
| klaus | rurban: I don't know, didn't spot it | ||
| rurban | Interesting. I'll have a look | ||
| klaus | the output of 'make test' is a bit noisy | 14:20 | |
| dalek | rrot: eaf4f48 | rurban++ | tools/release/cut.pl: Improve error reporting of tools/release/cut.pl for IO::Compress::Bzip2 If IO::Compress::Bzip2 will fail to load (Compress::Raw::Bzip2 version 2.059 required--this is only version 2.052) tools/release/cut.pl will write the bzip2 compress archive as gzip compressed. This might not be what you want. By adding "use IO::Compress::Bzip2;" the error is not silently ignored anymore. After updating Compress::Raw::Bzip2 the bzip compression worked fine. |
14:22 | |
|
14:23
davidfetter joined
|
|||
| rurban | sudo apt-get install libpcre3-dev freeglut3-dev libgmp-dev | 14:25 | |
| klaus | will try later | 14:26 | |
|
14:42
travis-ci joined
|
|||
| travis-ci | [travis-ci] parrot/parrot#717 (master - eaf4f48 : Reini Urban): The build was broken. | 14:42 | |
| [travis-ci] Change view : github.com/parrot/parrot/compare/3...f4f48b6b40 | |||
| [travis-ci] Build details : travis-ci.org/parrot/parrot/builds/3739854 | |||
|
14:42
travis-ci left
|
|||
| rurban | hmm, pmc.task.t still misbehaving... | 14:45 | |
| travis-ci.org/parrot/parrot/jobs/3739857 | |||
|
14:56
klaus_ joined
|
|||
| klaus | rurban: gist.github.com/4337237 | 15:01 | |
| rurban | hmm, not much infor | ||
| klaus | but can't spot that in the output of 'make test' | 15:02 | |
| rurban | ah, I see something similar now | 15:07 | |
|
15:07
klaus_ joined
|
|||
| rurban | It's the GC thread bug, with ./parrot examples/threads/chameneos.pir | 15:12 | |
| klaus | rurban: ? | 15:16 | |
| rurban | github.com/parrot/parrot/issues/880 + github.com/parrot/parrot/issues/875 | 15:17 | |
| dalek | rrot: 98351be | rurban++ | t/native_pbc/Test.pm: [t] todo native_pbc tests if only the version cmp fails pbc is now version-safe. mark the tests as TODO for a while. keep the tools/dev/pbc_header.pl --update-fingerprint hint. |
15:30 | |
| rrot: 549f84e | rurban++ | / (4 files): Remove outdated native_pbc files tools/dev/mk_packfile_pbc is wrong and not needed anymore. remove t/native_pbc/string.pbc and the references to t/native_pbc/integer.pbc, t/native_pbc/annotations.pbc. t/native_pbc/annotations.pbc was replaced by t/pmc/testlib/annotations.pbc, t/native_pbc/string.pbc by t/op/testlib/test_strings.pbc, t/native_pbc/number.pbc by t/pmc/testlib/number.pbc. t/native_pbc/*.pbc is for generated pbc files on different platforms, not on the same platform. |
|||
| rrot: ca694fb | rurban++ | docs/project/release_manager_guide.pod: [doc] Update release_manager_guide.pod for mk_native_pbc It is recommended to run pbc_header.pl --update-fingerprint, so that the native_pbc tests will not be TODOd. Remove the reference to the wrong tools/dev/mk_native_pbc. |
|||
| rrot: 4ea38b3 | rurban++ | t/native_pbc/ (11 files): Update versions in the native_pbc test pbcs Generated by pbc_header.pl --update-fingerprint on all PBC_COMPAT files (Skip the old 4.0 *8_be.pbc files) |
|||
|
15:52
dmalcolm joined
|
|||
| dalek | rrot/tt1589_library_path: b204751 | dukeleto++ | / (3 files): [TT #1589] Remove "." from default search paths INCLUDE, LIBRARY and LANG Apply patch from soh_cah_toa++ with small tweaks and an extra test. Before "." was at the top of the search paths, which implied a security risk. The other possibility from [GH #368] would be to add "." to the end of the search paths. Test that . is not in the library search path, and that runtime/parrot/dynext is searched. TODO: Deprecation, docs. |
15:53 | |
|
16:21
zby_home joined
16:40
benabik_ joined
16:42
benabik joined
17:12
Reini joined
17:32
benabik joined
|
|||
| benabik | ~~ | 17:38 | |
|
17:46
Reini joined
18:09
contingencyplan joined
18:18
ingy joined
|
|||
| rurban | oh, welcome ingy at #parrot | 18:25 | |
| I'm fed up with Library search paths [GH #368] lists.parrot.org/pipermail/parrot-d...07303.html | 18:26 | ||
|
18:29
ingy joined
18:46
Reini joined
19:17
TonyC joined
19:20
Reini joined
19:22
TonyC joined
19:31
TonyC joined
19:38
TonyC joined
19:43
TonyC joined
19:48
TonyC joined
19:56
Psyche^ joined
20:01
PerlJam joined
|
|||
| rurban | dalek: where are you? | 20:19 | |
| benabik | I think dalek is hosted on feather. | ||
| Which is currently MIA. | |||
| rurban | I see | 20:20 | |
| I've committed several updates to library_path-gh368 which is now ready | |||
| benabik | Fixes all your issues with library paths? | 20:21 | |
|
20:21
Reini joined
|
|||
| rurban | yes | 20:21 | |
| I even compare the expected output with eq: github.com/parrot/parrot/blob/6ce8...rch_path.t | |||
| benabik | Search paths can be altered inside the interp? | 20:23 | |
| rurban | interestingly yes, even deleted | ||
| I did some pop's | |||
| benabik | That's fairly useful. | 20:24 | |
| Although a security nightmare. | |||
| rurban | same as with perl5 | ||
| benabik | Although parrot in general isn't a very secure system. :-d | ||
| rurban | I guess a hardened libparrot should make the interp array read-only (deep) | 20:25 | |
| benabik | Well, hardening libparrot needs a lot of other things, so NBD? | ||
| rurban | Makes a nice GSoC project | ||
| Do you have time for the GC issues? | 20:26 | ||
| My idea is: GC can only run from the main interp, and can walk into all thread interps | |||
| benabik | Not thread safe. | 20:27 | |
| rurban | This way we can block alloc attempts from threads while walking the parent interp | ||
| benabik | objects could be altered in threads while main is doing GC, making the sweep invalid. | ||
| rurban | only block while mark | 20:28 | |
| benabik | From what I understand nine's design was that each thread sweeps their own objects. | ||
| You'd have to block the threads completely, not just alloc. | |||
| Lack of stop-the-world is a major advantage to current design. Current issues just seem to be bugs where GC is escaping into the wrong interp. Is better to fix the bugs instead of throw out the design. | 20:30 | ||
|
20:42
schmoo joined
20:45
schmoo joined
21:21
Reini joined
21:25
cooper joined
21:59
Reini joined
22:02
Reini1 joined
22:17
dalek joined,
p6eval joined
22:19
Coke joined,
PerlPilot joined
22:20
Util joined,
pmichaud joined
22:22
tadzik joined
22:28
Reini joined
22:58
donaldh joined
23:05
davidfetter joined
|
|||