|
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 August 2016. |
|||
|
01:48
ilbot3 joined
|
|||
| moderator | 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 | ||
| ugexe | why do we determine various identifiers based on the sha1 of names/paths instead of the sha1 of the contents of what those point at? Is it because nqp::sha1 would be too slow, or because of the added complexity of having a reference counter for each file (i.e two different versions have a file with the same sha1, so uninstalling one version shouldn't delete said file)? | 16:03 | |
| jdv79 | what's the benefits of uniquing on contents? | 18:11 | |
| TimToady | see git? | 18:12 | |
| specifically, it enforces immutability | 18:13 | ||
| jdv79 | i thought at the moment we were just using name hashing as a cheap fix for fs encoding issues | 18:15 | |
| TimToady | whereas while name/version/auth is mandated to be unique for officially installed modules, it doesn't specify how to enforce it | ||
| jdv79: almost nothing has a single reason in Perl 6 :) | |||
| jdv79 | it would be nice if the paths were hash free. its near impossible to figure out what files one is looking at when the editor tab just has a bunch of partial hash values | 18:16 | |
| it slows down debugging significantly for me | 18:17 | ||
| TimToady | maybe we could symlink human-readable names as guesses, and make collisions fail-soft | 18:18 | |
| jdv79 | the immutibility concern could be addressed outside of the filenames. maybe just in precomp and a side channel - maybe meta? | 18:19 | |
| not all fses support sylinks:( | 18:20 | ||
| *symlinks | 18:21 | ||
| anyway, ugexe that's my guess. no idea really. | |||
| mst | I suggested, ages ago, that we used IDN syntax for dealing with the weird naming stuff | 18:30 | |
| and I thought people had agreed that was a good idea | |||
| perhaps they agreed it was a good idea that I wrote the patch for that and I didn't realise | 18:31 | ||
| TimToady | :) | 18:40 | |
|
20:47
smls joined
|
|||
| smls | The zef install failure on `rakudobrew build moar-blead`, has returned for me... | 20:48 | |
| "Install [FAIL] for zef:auth('github:ugexe'): zef:ver<*>:auth<github:ugexe>:api<> already installed" | |||
| Is it an known issue? | |||
| ugexe | smls: `rakudobrew build moar-blead` does not attempt to reinstall zef for me | 22:32 | |
| smls | ugexe: Did you have it installed under moar-blead previously? | 22:33 | |
| ugexe | yes | ||
| smls | hm | ||
| ugexe | jdv79: another reason is deduplication, like how Dovecot dbox deduplicates attachments instead of saving every copy. this seems like a good fit for resources/, not so much for source code. | 23:00 | |
| the other current problem with duplication is `"provides" : { "Module::Foo::Bar" : "lib/Module/Foo/Bar.pm6", "Module::Foo::Bar::Helper" : "lib/Module/Foo/Bar.pm6" }` | 23:15 | ||
| which is ironically caused by the hashing (the above creates 2 files but uses the module name to hash and not the duplicated file name) | 23:17 | ||
| gist.github.com/ugexe/ddc78c5b39a1...68ead88c11 this is a fairly complete meta-level solution using sha1 filenames. unfortunately it doesn't really help its case | 23:23 | ||