| 04:58 | <synod> | Hi, where is the HTML source control repository? |
| 05:00 | <synod> | i did $ svn co http://svn.whatwg.org/webapps/ but it seems to be missing lots of things like the build system. |
| 05:15 | <synod> | spec build system? anyone? |
| 05:28 | <synod> | ok maybe build system for html is too hard... where are the parsing tests? |
| 05:43 | <synod> | no official html5 to xml tests? |
| 05:51 | <[tm]> | synod: build script isn't in the repo |
| 05:51 | <[tm]> | not really clear why anybody else would need it |
| 05:51 | <[tm]> | dunno what you mean by html5 to xml tests |
| 05:52 | <[tm]> | there's noththing that would make building something from html any harder than from xml |
| 05:53 | <synod> | [tm]: i would like to see how the spec is built... ? |
| 05:53 | <synod> | [tm]: as i understand it, html5 defines a way to parse invalid html into a valid dom? there is an xml version of each valid dom? |
| 05:56 | <[tm]> | xml version of each valid DOM... |
| 05:57 | <[tm]> | once the parser construcs the DOM it really doesn't matter where the serialized source it came from was text/html or XML |
| 05:58 | <synod> | how do i test my html5 parser? i have to diff the dom from html and the 'correct' dom, no? |
| 05:58 | <synod> | there are lots of xml -> dom parsers. are there html5 and xml test pairs? |
| 05:59 | <[tm]> | synod: http://www.whatwg.org/specs/web-apps/current-work/multipage/the-end.html#coercing-an-html-dom-into-an-infoset might have something you want as far as generating XML from a DOM, I dunno |
| 05:59 | <[tm]> | what is a a "correct" DOM man? |
| 05:59 | <[tm]> | why dod you want "test pairs"? |
| 05:59 | <[tm]> | what are you trying to do? |
| 06:00 | <[tm]> | there are lots of html -> dom parsers of course |
| 06:00 | <synod> | [tm]: i am wanting to write an html5 parser... how do i test it |
| 06:01 | <[tm]> | ah |
| 06:01 | <[tm]> | OK, well, there's a testsuite in the html5lib source repoo |
| 06:02 | <[tm]> | http://code.google.com/p/html5lib/ |
| 06:02 | <[tm]> | hg clone https://code.google.com/p/html5lib/ |
| 06:03 | <[tm]> | synod: what language you writing your parser in? |
| 06:03 | <synod> | [tm]: ocaml |
| 06:04 | <synod> | html5lib tests are in some custom format? |
| 06:04 | <[tm]> | python/html5lib/tests subdir |
| 06:05 | <[tm]> | the ones there are in python scripts |
| 06:06 | <[tm]> | there's json test sources in testdata subdir |
| 06:06 | <synod> | and some weird format with #data and ||| stuff |
| 06:07 | <[tm]> | yeah |
| 06:07 | <synod> | is html5lib tests the most complete for parsing? others look like rendering and such |
| 06:08 | <synod> | does html5lib track whatwg or w3c? |
| 06:08 | <[tm]> | none of them are rendering tests |
| 06:09 | <synod> | i mean the other suites linked from the wiki |
| 06:09 | <[tm]> | synod: there is no difference between the parsing algorithm in the whatwg and the w3c specs |
| 06:09 | <[tm]> | html5lib test suite is the most complete |
| 06:09 | <synod> | oh, that's good |
| 06:09 | <[tm]> | browser teams have upstreamed tests back to it |
| 06:10 | <[tm]> | hundreds of tests |
| 06:11 | <synod> | cool |
| 06:11 | <synod> | thx 4 the info |
| 06:12 | <[tm]> | synod: https://github.com/w3c/html-testsuite/tree/master/tests/submission/Opera/html5lib has a runnable harness that may or not be useful to you |
| 06:13 | <[tm]> | for more answers, hang out here for a while and ping jgraham and zcorpan and annevk when they wake up |
| 06:13 | <[tm]> | (CET time) |
| 06:16 | <synod> | ok |
| 06:38 | <annevk> | synod: you want to write a spec? |
| 07:11 | <synod> | annevk: for a different language |
| 07:11 | <synod> | annevk: looks like anolis is what whatwg uses? |
| 07:11 | <annevk> | http://wiki.whatwg.org/wiki/Anolis is what most of our specs use |
| 07:11 | <annevk> | in particular "Setting up Anolis with cross-specification cross-references" |
| 07:12 | <annevk> | HTML uses Anolis, but has several pre-processing steps of its own too |
| 07:13 | <synod> | ok... and html5lib tests are the most complete parsing suite? |
| 07:17 | <annevk> | yup |
| 07:22 | <synod> | thx... ttyl |
| 07:49 | <SimonSapin> | annevk: I skimmed the state machine in the URL Standard. Do you think that parts of it could be expressed (more tersely) as regular expressions, like RFC 3986 did? |
| 07:52 | <annevk> | if you do something like that you'd need to explain a whole bunch of other things |
| 07:55 | <annevk> | in general I don't think using ABNF or regular expressions is a good way to explain how to process complex syntax |
| 08:04 | Ms2ger | kicks annevk out of the IETF |
| 08:05 | <Ms2ger> | ... oh wait |
| 08:44 | <[tm]> | so I'm thinking about http://lists.w3.org/Archives/Public/public-html/2012Sep/0499.html |
| 08:44 | <[tm]> | "we simply ask browser vendors to provide |
| 08:44 | <[tm]> | the results they got when they regression tested the build that we are |
| 08:45 | <[tm]> | considering for conformance purposes" |
| 08:45 | <[tm]> | which sounds worth considering |
| 08:45 | <[tm]> | but would still want to be able to indepedently verify the results if possible |
| 08:46 | <[tm]> | I wonder if all browser teams have test runners that I could set up instances of |
| 08:48 | <Ms2ger> | Can you set up instances of Microsoft employees? |
| 09:11 | <[tm]> | Ms2ger: :-) |
| 14:43 | <jgraham> | [tm]: You are quite welcome to verify the results if you like, but I think it will take you a while :) I'm not sure that we would be desperate to go through the effort of releasing our test infrastructure |
| 14:48 | <jgraham> | Not because I see it as secret sauce, just because it would be quite a lot of work to unleash a mess of code that was designed to run in our particular environment on the world, and very little gain |
| 15:54 | <jgraham> | Is it just my computer or is clang/llvm roughly the slowest thing in the world to compile |
| 15:54 | <Ms2ger> | Nah |
| 15:54 | <Ms2ger> | Gecko is |
| 15:54 | <jgraham> | (I mean to compile clang itself, not to compile using clang) |
| 15:55 | <jgraham> | Well I think either this computer is more rubbish than I believed or I am doing something seriously wrong because so far this seems to have taken 3 hours |
| 15:58 | <Ms2ger> | bz says 15 minutes at most |
| 15:58 | <[tm]> | jgraham: thanks for comments about test stuff |
| 15:59 | <Ms2ger> | But he's got pretty fast hardware |
| 15:59 | <smaug____> | yeah, I don't think compiling clang is that slow |
| 15:59 | <zewt> | well, unless there's a huge RAM issue, the difference between crappy budget hardware and really fast consumer hardware isn't 15-minutes/3-hours |
| 16:00 | <smaug____> | swapping would make compiling really slow |
| 16:00 | <zewt> | that's a "huge RAM issue" :) |
| 16:03 | <jgraham> | It could well be. It has certainly swapped everything else out |
| 16:04 | <jgraham> | But is showing free memory so... |
| 16:05 | <zewt> | i don't even use swap on any of my linux systems, heh, but i don't run X on most of them |
| 16:05 | <jgraham> | This is OSX |
| 16:05 | <jgraham> | Which is super-annoying when you want to compile things |
| 16:06 | <jgraham> | Or really install anything |
| 16:06 | <zewt> | osx is super-annoying when you want to ... things |
| 16:07 | <zewt> | (keyboard navigation for menus? hahahaha) |
| 16:17 | <Ms2ger> | Mm, I think there's a clang in homeports or brews or whatever |
| 16:45 | <jgraham> | zewt: Well it's not like I plan to buy a Mac again. Would be nice if someone else was making comparably good hardware though |
| 16:45 | <jgraham> | Well specifically comparably good ouchpads |
| 16:45 | <jgraham> | *touchpads |
| 16:47 | <[tm]> | jgraham: I built llvm on my OS X laptop before and I am certain it took less that 3 hours |
| 16:48 | <[tm]> | from the sources, with defaults -- debug build plus all whatever test stuff and such it also builds |
| 16:48 | <[tm]> | so unless something major has changed in the sources since then, I wouldn't think it should take that long |
| 16:49 | <[tm]> | that said, I'm also sure it took more than 15 minutes |
| 16:49 | <[tm]> | it wasn't fast |
| 16:49 | <[tm]> | I'd guess it was probably 30 minutes or definitely less than hour |
| 16:49 | <[tm]> | fro llvm+clang |
| 16:51 | <[tm]> | 2.7 GHz dual core |
| 16:52 | <[tm]> | it was faster than building some other stuff -- e.g., boost |
| 17:36 | <zewt> | random sad hilarity: tabbing to the search box in Steam with the keyboard, instead of clicking, lands you on an input box containing the text "search the store" |
| 22:31 | <teear> | hello |