Fire is step THREE! | github.com/perl6/toolchain-bikeshed | Channel logs: irclog.perlgeek.de/perl6-toolchain/today | useful prior art: metacpan.org/pod/CPAN::Meta::Spec
Set by moderator on 11 May 2016.
05:19 TimToady joined 05:55 domidumont joined 06:00 domidumont joined
nine MadcapJake: my refactoring is mostly done :) It should be reasonably safe now to write a CompUnit::Repository. 07:49
Having an alternative CompUnit::PrecompilationStore has become more possible, too now. I moved much of the file access into CompUnit::PrecompilationStore::File. But there are still a couple of loose ends. 07:51
However I think the most pressing issue is now to finally get ugexe's Distribution overhaul merged and get a much more reasonable interface for installers. 07:52
12:45 cognominal joined 13:46 cognominal joined 14:19 sufrostico joined
ugexe i can go over things again and fix conflicts, but we should probably figure out a few details first. 1. what/how to break meta spec (i.e. auth/authority/author) 2. special handling of version * (if something with version '*' gets installed you cannot install that module with any other version without force). Maybe when installing version '*' in the meta should mean A. installed but always update or 14:27
B. installed but only *string* matches '*' (again, only for meta and during installation) 3. CURI API or something to get the revdeps to uninstall distributions 4. Figure out a way/api for resources to install non-mangled names (seems to be OS specific cases, so these things should already be named safely: see windows dlls that cant be renamed)
to clarify point 4 see this commit to openssl that was needed because nativecall said it could not find the dll with the mangled name (the real error must be something else, because the file actually exists) github.com/sergot/openssl/commit/9...4cfedbfa9f 14:30
jdv79 oh lovely. nqp::sha1 being used outside. should be fun to change that to a human readable form then. 14:33
ugexe well if you ever get a change to look in a windows $*TMPDIR you'll see it fits right in 14:35
but yeah its a giant hack that needs to die asap
nine I'd totally be up for 123456789abcdef/foo.so instead of 123456789abcdef.so 14:38
To be safe we could use punycode for the file name. foo.so should stay untouched this way and we're still fine with Unicode. 14:39
Using Unicode dll names and expecting it to work on Windows will just be too much
ugexe why does renaming the dll affect it anyway? you can move the dll to different folders just fine, so i can only guess the base name gets baked into the symbol names or something? 14:41
Cannot locate native library 'C:\\rakudo\\share\\perl6\\site\\resources\\5AC8B407FC19FF40B94E90605A758D6AF872EF83.dll': error 0x7f in method setup at sources\\24DD121B5B4774C04A7084827BFAD92199756E03 (NativeCall) line 279 # this is the exact error; the file mentioned does actually exist 14:45
nine Maybe it's not even Windows but NativeCall that tries to be smart somehow? 14:52
14:56 hankache joined
ugexe could be. i traced it to nqp::buildnativecall (which appears to get passed all the correct arguments) but looking at that code was greek to me 14:56
nine With my latest commits we should be good for building as unpriviledged user and installing system wide now. 15:28
Successfully created rpm packages that work :) 15:29
jdv79 is punycode better than simple uri style encoding of higher than ascii? 15:38
i think that's what CHI uses and i believe its cross platform
nine Oh yes, that may work just as well 15:42
jdv79 i'm sure there is way better material but this is what i remember from last $work: github.com/jonswar/perl-chi/blob/m...#L737-L742
fairly boring 15:43
nine I'd say any simple encoding that leaves plain ASCII intact is good
jdv79 yeah. punycode just seems uncommon to me but maybe that's just me. i know uri-ish encoding pretty well cause its very common (again to me)...:) 15:44
almost anything is better than the opaqueness that exists now
nine Though I've read much fewer complaints since backtraces are readable again 15:46
mst honestly, people putting clever things in file names can enjoy whatever happens to them 15:56
I just want 'locate' to work rather than the mad sha thing ;)
I like punycode better than URL-ish encoding, I think 15:57
for many reasons, but 'fits into DNS sort of for people wanting java-style fdqn based hierarchies' strikes me as neat
16:30 domidumont joined 16:38 domidumont1 joined
jdv79 what is the dns-ness valueable for? does a dns parser of some sort need to be able to parse namespaces? 17:14
mst jdv79: org.people.kill.me.now type package naming 17:15
does have its value
jdv79 and where is punycode's advantage? is that actually used in a dns-ish way? i know next to nothing about the finer details of java or that naming scheme. 17:17
17:17 Kassandry joined
jdv79 well, ok. i guess. yeah, anything but ascii doesn't need to be super easy to read i guess. 17:18
though i would still argue against punycode simply on grounds of human readability. 17:29
mst Punycode is used to encode internationalized domain names
jdv79 can you decode it in your head? 17:30
[Coke] jdv79: punycode is more readable than url encoding. YMMV.
punycode++ from me. 17:31
jdv79 maybe i just haven't seen enough of it then
[Coke] Not for detranslating, necessarily. Here's an example: 17:32
jdv79 sounds like we have a quasi-winner then
[Coke] minimal ünicode -> xn--minimal nicode-nsb
17:34 domidumont joined
mst that requires a lot less escaping to paste around than URL encoding IME 17:36
jdv79 cool. thanks for the help with that. 17:40
18:11 TimToady joined
lizmat ping 19:23
nine pong 19:27
lizmat :-) 19:32
20:50 patrickz joined
jdv79 ding 20:51
21:17 sufrostico joined 23:35 sufrostico joined 23:46 autarch joined

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!