| 12:10 | <hsivonen_> | Does anyone happen to have a demo showing how document.write from a timeout during the parse behaves in IE? |
| 12:10 | <hsivonen_> | it blows away the document, right? |
| 12:13 | <zcorpan_> | i'm pretty sure it does |
| 12:13 | <hsivonen_> | thanks |
| 12:17 | <zcorpan_> | hsivonen_: did you see the script bug? |
| 12:17 | <hsivonen_> | zcorpan_: I didn't. What's the bug? |
| 12:18 | <zcorpan_> | hsivonen_: i sent an email. <script><!--<script></script>x</script>y |
| 12:18 | <zcorpan_> | the </script>x does not appear in the dom |
| 12:18 | <hsivonen_> | zcorpan_: thanks |
| 12:19 | hsivonen_ | sees the PFWG has sent out responses to ARIA review comments |
| 12:19 | hsivonen_ | needs to review the comments |
| 12:19 | <hsivonen_> | s/comment/responses/ |
| 12:23 | <annevk2> | yeah me too |
| 12:23 | <annevk2> | at some point |
| 12:23 | <annevk2> | too much work :/ |
| 12:23 | <hsivonen_> | zcorpan_: I think the bug you found is the same as https://bugzilla.mozilla.org/show_bug.cgi?id=533111 |
| 12:24 | <hsivonen_> | zcorpan_: your test case is much nicer :-) |
| 12:26 | <zcorpan_> | hsivonen_: looking at comment 3, it appears to be different |
| 12:33 | hsivonen_ | wonders what possesses Web authors to call document.write() from timeouts |
| 13:30 | <hsivonen> | gsnedders: is html5lib missing test coverage for the new script states? |
| 13:30 | <gsnedders> | hsivonen: No |
| 13:30 | <gsnedders> | hsivonen: Or at least, not much |
| 13:30 | <hsivonen> | gsnedders: hmm. I don't see script tests in escapeFlag.test or in contentModelFlags.test |
| 13:31 | <gsnedders> | hsivonen: test16.dat |
| 13:31 | <hsivonen> | gsnedders: ah. excellent. thanks |
| 13:32 | <gsnedders> | hsivonen: Complains about missing tests should be directed at zcorpan |
| 13:33 | <hsivonen> | eek. I see lots of test failures |
| 13:33 | <gsnedders> | In html5lib? |
| 13:34 | <hsivonen> | in V.nu |
| 13:34 | <gsnedders> | Or in validator.nu? |
| 13:34 | <gsnedders> | (html5lib test failures are caused by a lot of tests not having the correct expected parse errors) |
| 13:34 | <hsivonen> | so I'm noticing |
| 13:35 | gsnedders | turned that on as it turned up several parsing bugs |
| 13:35 | <hsivonen> | is it ok to push test fixes without fixing html5lib itself? |
| 13:35 | <gsnedders> | Yeah sure |
| 13:35 | <hsivonen> | ok |
| 13:51 | <zcorpan> | shouldn't a loop in structured clone throw INVALID_STATE_ERR instead of NOT_SUPPORTED_ERR? |
| 13:52 | <Philip`> | Loops aren't invalid, they just aren't yet supported by the algorithm |
| 13:55 | <zcorpan> | Hixie: ^ |
| 13:56 | <annevk2> | what's wrong with loops? |
| 13:57 | <zcorpan> | it's hard to get a clone |
| 13:58 | <annevk2> | but technically you can clone them I think |
| 13:59 | <Philip`> | It's not that hard to implement, you just change the 'memory' list so that for each object you store a reference to its clone, and if you're cloning something that's already in the memory then you copy the stored reference |
| 14:19 | <zcorpan> | should we change the algorithm to support loops? |
| 14:22 | <Philip`> | (If it supports loops, it should also support x=[]; y=[x,x]; clone(y) without breaking the y[0]==y[1] equality) |
| 14:22 | <Philip`> | (which is different to the current behaviour) |
| 14:34 | <boblet> | hey Mike |
| 14:43 | <MikeSmith> | boblet: you're definitely welcome to stay at my place |
| 14:43 | <zcorpan> | seems like that's the current behavior in our impl |
| 14:43 | <boblet> | MikeSmith: thanks. sorry for the late notice (!) |
| 14:44 | <MikeSmith> | np |
| 14:44 | <MikeSmith> | don't need to bring anything |
| 14:45 | <boblet> | ok. see you at the event then |
| 14:47 | <MikeSmith> | boblet: yep, see you there |
| 14:50 | <MikeSmith> | zcorpan: I've been looking at implementing xml-sheetsheet PI checking (bug 14) |
| 14:51 | <MikeSmith> | talking with Yudai about it too |
| 14:52 | <MikeSmith> | Yudai is spending some time part-time as an intern and sysadmin at W3C/Keio |
| 14:52 | <zcorpan> | MikeSmith: cool |
| 14:53 | <MikeSmith> | zcorpan: one question, for cases like <?xml-stylesheet foo bar?> |
| 14:53 | <MikeSmith> | what error should we emit? |
| 14:54 | <zcorpan> | MikeSmith: an error with a message that says the pi failed to parse |
| 14:56 | <zcorpan> | MikeSmith: from the spec's perspective, it's a potential xml-stylesheet pi whose parsing results in an error |
| 14:57 | <MikeSmith> | zcorpan: understood. but I'm not sure that's the most helpful thing to say to end uses |
| 14:57 | <MikeSmith> | *users |
| 14:57 | <zcorpan> | MikeSmith: or rather, it's a pi with target xml-stylesheet that's not an xml-stylesheet pi (because it failed to parse) |
| 14:57 | <MikeSmith> | zcorpan: what I have it emitting now is, "Potential xml-stylesheet processing instruction is ill-formed." |
| 14:57 | <MikeSmith> | I guess I could make that "Potential xml-stylesheet processing instruction failed to parse." |
| 14:58 | <zcorpan> | if you could say why it failed to parse, that's probably be more helpful |
| 15:00 | <zcorpan> | MikeSmith: the message from the xml parser in v.nu is "required character (found b) (expected =)" |
| 15:01 | <zcorpan> | for <html foo bar> |
| 15:09 | <hsivonen> | Hixie: is there any chance of getting dreamhost to fix the gzip use on lists.whatwg.org for range requests? |
| 15:23 | <Philip`> | "Can sites host their own copy of foaf.html to avoid failure in communication when example.org is down?" - hmm, sadly http://uptime.netcraft.com/up/graph?site=example.org doesn't seem to give uptime data for it |
| 15:31 | <TabAtkins> | I am the luckiest man alive. I won a TV on Friday, and a pie this morning. I'm looking for a lotto ticket next. |
| 15:32 | <jgraham> | You might be hit by a bus. That is also a rather unlikely event so presumably indicates luck |
| 15:32 | <TabAtkins> | Nonono, good luck and bad luck are clearly separate things. |
| 15:33 | <zcorpan> | but both are still luck, no? |
| 15:33 | <TabAtkins> | Only in the abstract sense. They're not commensurable unless you believe in karma. |
| 15:35 | <Philip`> | Regardless of whether it counts as luck, you would no longer be the luckiest man alive |
| 15:35 | <jgraham> | You have to hope that people who believe in karma don't get lucky and get you as a textbook example of karma in action |
| 15:35 | <TabAtkins> | Philip`: No jinxing. >_< |
| 15:35 | <jgraham> | Philip`: Depends on the effects of the bus |
| 15:37 | <Philip`> | Well, I suppose it could be a magic bus which transforms into a field full of sunflowers when it hits somebody |
| 15:38 | <annevk2> | not all the ARIA replies seem satisfactory |
| 15:38 | <annevk2> | e.g. defining hidden element as done here http://www.w3.org/WAI/PF/aria/20091214/terms#def_hidden does not really seem implementable |
| 15:41 | <zcorpan> | hmm, definition reference to wcag2 |
| 15:42 | <annevk2> | and with respect to my datatype comment the reply says it depends on the host language... |
| 15:43 | <annevk2> | so you need a separate ARIA attribute parser for SVG and HTML? |
| 15:43 | <annevk2> | and for any new language? |
| 15:49 | gsnedders | writes nonsense in his email to public-html |
| 15:52 | <MikeSmithX> | zcorpan: so the initial implementation of xml-stylesheet PI checking that I tried out involved internally transforming the PI into a document/element and then parsing that |
| 15:52 | <zcorpan> | MikeSmithX: ok |
| 15:52 | <MikeSmithX> | that was kind of what you suggested |
| 15:53 | <zcorpan> | yeah |
| 15:53 | <MikeSmithX> | and for that in the ill-formed pseudo-attribute case, Xerces says: 'Attribute name "foo" associated with an element type "xml-stylesheet" must be followed by the ' = ' character.' |
| 15:53 | <MikeSmithX> | which I guess is roughly same as your "required character (found b) (expected =)" message |
| 15:54 | <MikeSmithX> | zcorpan: anyway, for the actual end-user message, I think I will just go with "...failed to parse." |
| 15:54 | <MikeSmithX> | for that case |
| 15:55 | <zcorpan> | yeah, that's probably good enough |
| 15:56 | <MikeSmithX> | btw, I thought about the idea that you suggested of having an xml-stylesheet PI mini-schema and then validating the instances against that |
| 15:56 | <MikeSmithX> | but |
| 15:56 | <MikeSmithX> | the problem with that would be that the error messages would call the pseudo-attributes "attributes" |
| 15:57 | <MikeSmithX> | which they aren't of course |
| 15:57 | <zcorpan> | can't you post-process the message? |
| 15:57 | <MikeSmithX> | no, not easily |
| 15:58 | <MikeSmithX> | if we could there would be a lot of other places where it'd already be doing it |
| 15:59 | <MikeSmithX> | it's one of the tradeoffs in using an off-the-shelf general validation mechanism |
| 15:59 | <MikeSmithX> | as opposed to doing all the checking with custom code |
| 15:59 | <MikeSmithX> | Henri talks about it in his thesis |
| 15:59 | <MikeSmithX> | anyway, it's not a big problem in this case |
| 16:00 | <MikeSmithX> | I'm mostly just being lazy |
| 16:01 | <MikeSmithX> | and trying to figure out the easiest way to parse the set of pseudo-attributes without needing to write specific parsing code to do it |
| 16:01 | <zcorpan> | i guess it's not so bad if the messages say "attribute" instead of "pseudo-attribute" |
| 16:02 | <MikeSmithX> | zcorpan: maybe, but not sure hsivonen would agree about that |
| 16:04 | <zcorpan> | MikeSmithX: what happens currently for pis in the internal subset? |
| 16:04 | <MikeSmithX> | zcorpan: I don't know because I've not looked at the xmlparser code much |
| 16:05 | <MikeSmithX> | zcorpan: but I think the xmlparser code is mostly just doing whatever aelfred does |
| 16:05 | <zcorpan> | MikeSmithX: do you have a running prototype? |
| 16:06 | <MikeSmithX> | zcorpan: I can have it up on qa-dev shortly |
| 16:06 | <MikeSmithX> | well, relatively shortly |
| 16:06 | <MikeSmithX> | I need a break first |
| 16:07 | <zcorpan> | ok |
| 16:07 | <MikeSmithX> | after two hours of phone conferences |
| 16:07 | <MikeSmithX> | but I will ping you again when I have it up |
| 17:56 | <MikeSmithX> | if zcorpan comes back around, please let him know that http://qa-dev.w3.org:8888/ now has a minimal xml-stylesheet PI checker |
| 17:58 | <MikeSmithX> | it doesn't report anything yet except the simplest case, which is if code fails to parse the PI at all |
| 18:32 | <gsnedders> | 'The "rows' and "cols" attributes aren't presentational, they give the wrapping edge of the textarea, which affects submission' — how? |
| 18:33 | <gsnedders> | Based upon the wrap attribute? |
| 18:33 | <gsnedders> | So it's only something new in HTML 5 that has any affect? |
| 18:58 | <Philip`> | gsnedders: wrap isn't new in HTML5 |
| 18:58 | <Philip`> | It's existed at least since IE4 |
| 18:59 | <gsnedders> | Ah |
| 19:46 | <AryehGregor> | gsnedders, well, cols is used for wrap=hard. Doesn't seem rows is used for anything. |
| 19:47 | <AryehGregor> | Does CSS have a good way to emulate them? Would you have to manually multiply the font-size times the line-height or something? |
| 19:48 | <AryehGregor> | That's just for rows, cols I have no idea on (but it's maybe less important, since textareas are more often wide than tall, and in a wide textarea it's not so noticeable if you have some extra empty pixels at the side) |
| 19:48 | <TabAtkins> | font-size times line-height? ?_? Just line-height times the number of rows you want. |
| 19:49 | <TabAtkins> | (Unless you are expressing your line height relative to your font-size, of course. |
| 19:49 | <TabAtkins> | cols are much less important, but you can typically estimate it with ems. |
| 19:50 | <AryehGregor> | Hmm, true, but you still have to manually check what that's supposed to be, right? |
| 19:50 | <AryehGregor> | The line-height, I mean. |
| 19:50 | <AryehGregor> | You can't just use ems or something, in other words. |
| 19:51 | <TabAtkins> | Yes. |
| 19:51 | <TabAtkins> | If your line-height is the standard 1.2, you'll have to do height:(1.2 * number of rows)em; |
| 19:52 | <TabAtkins> | At least, that should approximately work. |
| 19:52 | <AryehGregor> | Wow, Mozilla has a heck of a lot of bugs. 211 results for "textarea". |
| 19:53 | <AryehGregor> | Hmm, this looks relevant: https://bugzilla.mozilla.org/show_bug.cgi?id=502462 |
| 19:53 | AryehGregor | has noted that <textarea maxlength> doesn't work in Firefox |
| 19:54 | <TabAtkins> | Nah, that bug's explanation is "<textbox multiline> doesn't work with maxlength because it's an html:textarea, which doesn't work with maxlength". |
| 19:55 | <AryehGregor> | That implies something to do with html:textarea not supporting maxlength, which is promising. |
| 20:05 | <zcorpan_> | MikeSmithX: i don't get any messages regarding for <?xml-stylesheet foo bar?> |
| 20:18 | <csarven> | Are localStorage keys sandboxed to the hosts in which they are created at or can they be shared? |
| 20:19 | <csarven> | Apparently they are |
| 20:37 | <TabAtkins> | Heh, there's nothing like demonstrating a working prototype of a proposal to make it really obvious why it's needed. |
| 20:40 | AryehGregor | is somewhat surprised that he was able to hack a trivial feature into Firefox in, like, an hour |
| 20:41 | AryehGregor | should try hacking on other projects more often |
| 20:49 | <mpt> | AryehGregor, what OS(es) do you use? |
| 20:49 | <AryehGregor> | mpt, Linux mainly, but I have Windows lying around too. |
| 20:50 | <AryehGregor> | (Ubuntu specifically) |
| 20:50 | <mpt> | ah |
| 20:50 | <AryehGregor> | Why do you ask? |
| 20:51 | <mpt> | We could always use some help with the Ubuntu Software Center. :-) Especially if you know your way around WebKit |
| 20:51 | <AryehGregor> | I don't, I'm a MediaWiki developer. |
| 20:51 | <AryehGregor> | If I were to branch out it would probably be into browser engines, though. |
| 20:53 | <mpt> | ok |
| 21:29 | <TabAtkins> | AryehGregor: What did you hack? |
| 21:37 | <virtuelv> | <offtopic> |
| 21:37 | <virtuelv> | I don't understand the software center |
| 21:38 | <virtuelv> | I install stuff once, and mostly won't touch the installation after that |
| 21:55 | <AryehGregor> | TabAtkins, I tried adding maxlength support to textarea. The patch is five lines and seems to work. |
| 21:56 | <Philip`> | AryehGregor: You can't fit test cases inside five lines |
| 21:56 | <Philip`> | so you need to do more patching :-p |
| 21:56 | <AryehGregor> | Yes, I know that part is missing. |
| 21:56 | <AryehGregor> | I have to figure that out. |
| 21:56 | <AryehGregor> | I'll have to look at what kinds of tests there are for maxlength on inputs. |
| 21:57 | Philip` | would not be entirely surprised if there weren't any |
| 21:57 | <Philip`> | (but nor would I be surprised if there were) |
| 21:57 | <AryehGregor> | I looked at some recent commits and not all of them seemed to have tests attached. |
| 21:59 | <Philip`> | I think there weren't really any tests at all until a very small number of years ago |
| 21:59 | <Philip`> | and it's unlikely anyone would have bothered adding maxlength tests since then, unless as part of recent bug fixes to it |
| 22:04 | <AryehGregor> | The code ends up being just "enable the input maxlength implementation for textareas", as far as I can tell, so maybe it doesn't really need independent tests. Perhaps I should just submit the patch and ask if it needs tests. |
| 22:06 | <Philip`> | AryehGregor: Everything should have tests :-) |
| 22:07 | <Philip`> | (That's probably not Mozilla policy, though) |
| 22:07 | <AryehGregor> | MediaWiki has an excellent test suite. However, it's not automated. You see, we use a wiki-based approach to testing: enable it on Wikipedia and wait for people to complain. |
| 22:07 | <TabAtkins> | !_! What the crap, Firefox? It's for some reason wrapping my <thead> across three rows in the table. |
| 22:07 | <Philip`> | (but it's common sense :-p ) |
| 22:08 | <Philip`> | (Otherwise someone will break the feature and nobody will notice) |
| 22:08 | <AryehGregor> | Obviously then it wasn't an important feature anyway. |
| 22:08 | <AryehGregor> | Hardly worth the effort. |
| 22:09 | <Philip`> | Nobody noticed when Firefox completely broke canvas drawImage, until after it was released |
| 22:09 | <AryehGregor> | I know a programmer who works on bond-trading software. He said he's had integer overflow when putting a sum of money in a 32-bit int. And he doesn't use automated test cases. So clearly some stupid *wiki* doesn't have to. |
| 22:10 | <AryehGregor> | :) |
| 22:10 | <AryehGregor> | See, that's one nice thing about developing MediaWiki, we barely have releases and don't care much about them. |
| 22:10 | <AryehGregor> | Break something? Bug a sysadmin and get the fix synced to Wikipedia, no need to worry about anyone else unless it's a security problem. |
| 22:11 | <Philip`> | I guess it's easier to deploy updates when the software is on a handful of servers rather than on a hundred million users' PCs, so you don't need to be as careful |
| 22:12 | <AryehGregor> | Well, Mozilla can push out updates too if they feel like it. |
| 22:13 | <AryehGregor> | Things like layout engines are probably a lot more complicated and fragile than most code, though. |
| 22:13 | <AryehGregor> | And of course they actually have to worry about competition. |
| 22:13 | <AryehGregor> | :) |
| 22:14 | <Philip`> | I think it's not just a case of "feel like it", given the time and effort and bandwidth and user irritation involved in pushing updates :-) |
| 22:27 | <TabAtkins> | Wow. Check out http://www.zachstronaut.com/posts/2009/02/17/animate-css-transforms-firefox-webkit.html in Firefox. The text in the middle box constantly dances. |
| 22:30 | <AryehGregor> | Firefox bookmarklet from blog.johnath.com: javascript:document.body.style.MozTransform="rotate(90deg)";void(0); |
| 22:30 | <AryehGregor> | Completely pointless, but amusing! |
| 22:30 | <AryehGregor> | javascript:document.body.style.WebkitTransform="rotate(90deg)";void(0); works for WebKit. |
| 22:35 | <othermaciej> | all the browsing gangstas hold their web pages sideways |
| 22:36 | <roc> | AryehGregor: for something that's easy to test automatically, like <textarea maxlength>, the reviewer would demand tests, I think |
| 22:37 | <AryehGregor> | roc, okay, but what should I test? The existence of .maxLength is easy to do in just JS, I guess. What sort of test would I have to write to test "it doesn't let me type more than X characters", a Mochitest? |
| 22:37 | <TabAtkins> | othermaciej: Hahaha. |
| 22:38 | <AryehGregor> | roc, also, should I request review from you, or is there anyone else specific you'd recommend? |
| 22:38 | <roc> | patches that land without tests are generally either not supposed to be changing behavior (although some of those have tests anyway), or are changing something that's very hard to test in a robust manner (e.g. exactly what pixels end up on the screen due to some bizarro Win32 windowing API call), or are security bugs and the test is an exploit |
| 22:38 | <AryehGregor> | Okay. |
| 22:39 | <roc> | I can probably review it |
| 22:39 | <roc> | although it might be better to get jst⊙mo |
| 22:40 | <roc> | you probably want to write a mochitest and use EventUtils.js with waitForFocus and synthesizeKeyEvent |
| 22:40 | <roc> | (and thanks!!!) |
| 22:40 | <AryehGregor> | :) |
| 22:44 | <TabAtkins> | Argh why is it so impossible to create vertical table headers. >_< |
| 22:49 | <Hixie> | http://www.w3.org/mid/OF3B70DB4D.2F7BF1C9-ON8625768D.0078BE03-8625768D.007A8C21⊙uic may be interesting to implementors and authors interested in <canvas> |
| 22:49 | <Hixie> | (please send feedback straight to that list) |
| 22:57 | <zcorpan> | Hixie: doesn't the device example need <video autoplay>? |
| 23:01 | <Hixie> | zcorpan: probably wouldn't hurt, indeed |
| 23:14 | <fantasai> | Anyone know why html5lib's documentation page returns 404? |
| 23:28 | <Hixie> | i just checked in r4444 |
| 23:28 | <Hixie> | four fours! it's gonna be a long time until r55555 |