|
08:36
lizmat_ joined,
TempIRCLogger__ left
08:37
lizmat left,
TempIRCLogger joined,
lizmat_ left,
lizmat joined
09:48
patrickb joined
|
|||
| JRaspass | lizmat: So that actually is 0.9.01 but the version in the meta is wrong: `"version" : "0.9.0"`. The 0.9.0 tarball isn't indexed because like you said the meta is invalid json. Not sure what we can do here when filename and meta disagree, we trust the meta. | 10:16 | |
| lizmat | hmmm.. ok, so you could argue the META in 0.9.01 is also wrong :-) | 10:17 | |
| it just happened to be valid JSON | |||
| wondering whether we shouldn't reject those cases also | 10:18 | ||
| hmmm... | |||
| JRaspass | i suspect the new eco would? | ||
| cpan is just a crapshot :-P | |||
| lizmat | true | ||
| fwiw, I think the version encoded in the distribution name should be used | 10:30 | ||
| because PAUSE would not allow uploads of the same version encoded in the distribution name, but it *would* allow uploads with a non-matching version in the META. | |||
| JRaspass | can't people put any old crap in the filename though? name-with-hyphens-version.tar.gz is just convention? | 10:31 | |
| lizmat | that is also true, but that would imply evil intent | ||
| it has happened to me that I've uploaded a new version without updating the version in the META | 10:32 | ||
| (prior to using App::Mi6 :-) | |||
| JRaspass | isn't the meta the primary source of truth though? | 10:33 | |
| and yeah you're not the only one who hasn't bumped | |||
| i could maybe see to trusting filename on version only i suppose, it would actually fix a lot of collisions atm where we take the latest only | 10:34 | ||
| lizmat | well in the case of CPAN, I'd say that wrt to version, the primary truth is indeed in the filename :) | ||
| JRaspass | im leaning slightly that way, ill try and run some numbers in a bit, i wonder how many dists disagree between filename and meta | 10:35 | |
| in the mean time i have a plan to finally allow crazy versions with letters, which should fix your uniname-words with its v6.c | |||
| lizmat | quite a few, it seems | 10:36 | |
| ah, no, my mistake | |||
| JRaspass | ah FROGGS has a v in the filename, that's rare - `Inline-v1 Inline-v1.1Inline-v1.2 Inline-v1.2.1` | ||
| lizmat | yeah, there are others | ||
| Vikna e.g. | |||
| jjatria | So if META is correct but filename is not, we ignore the META? | 10:37 | |
| That seems off | |||
| JRaspass | we ignore one key of the meta i'd say, just version | ||
| i feel dist sites are just hacks upon hacks to cope with the real world, i wonder how bad metacpan's code is :-P | 10:38 | ||
| lizmat | Sparky is a case, where there look to be at least 4 versions prior to 0.0.30 with incorrect version in META | ||
| JRaspass | yeah i don't think it's that rare | 10:39 | |
| jjatria | I understand why this hack is attractive, but I'm not a fan. I'd lean more towards to not indexing incorrect METAs. I think MetaCPAN also has a "not indexing?" page with some common causes | ||
| lizmat | gist.github.com/lizmat/178da597106...d151d38aea | 11:12 | |
| list of 70 distributions with version mismatch between META and distribution name | 11:13 | ||
| only a few of them because of the "v" in the version | 11:15 | ||
|
14:45
patrickb left,
patrickb joined
14:49
patrickb left,
patrickb joined
14:51
patrickb left,
patrickb joined
14:53
patrickb left,
patrickb joined
14:55
patrickb left,
patrickb joined
14:57
patrickb left
14:58
patrickb joined
15:04
patrickb left,
patrickb joined
15:06
patrickb left,
patrickb joined
15:08
patrickb left
15:09
patrickb joined
15:10
patrickb left
15:11
patrickb joined
15:13
patrickb left,
patrickb joined
15:15
patrickb left,
patrickb joined
15:19
patrickb left,
patrickb joined
15:23
patrickb left
|
|||