| 00:06 | <roc> | paul_irish: I believe CORS was first supported in Firefox 3.5. If not that, definitely 3.6 |
| 00:08 | <paul_irish> | roger that. thx roc |
| 00:35 | <TabAtkins> | Why do I torture myself by continuing to use http://www.xanthir.com/pictures/llama.gif as my example picture when playing with images? |
| 01:46 | <divya> | I am curious why almost all CSS examples in whatwg has a @namespace defined? http://www.whatwg.org/specs/web-apps/current-work/complete.html#margins-and-padding |
| 01:46 | <annevk> | those are not examples |
| 01:47 | <annevk> | and it makes sure those styles only apply to HTML elements and not e.g. XML elements that happen to share the name |
| 01:47 | <divya> | oops. right. sorry. |
| 01:53 | <JonathanNeal_> | What do you think of: "ARIA (Accessible Rich Internet Applications) is a collection of metadata added to or implied in HTML for the purpose of increasing the accessibility of content and components." ? |
| 01:54 | <JonathanNeal_> | And: "Microdata is a collection of metadata added to HTML for the purpose of conveying contact information, geographic coordinates, calendar events, and the like." ? |
| 02:06 | <divya> | JonathanNeal_: microdata defn sounds a lot like microformats one |
| 02:06 | <divya> | http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#introduction-2 is a bit more broad. |
| 02:06 | <JonathanNeal_> | Well, they are very similar, I botched together a bunch of web definitions to come up with a quick definition. |
| 02:07 | <JonathanNeal_> | I'll check it out right now, thanks |
| 03:33 | <annevk> | omg what a bunch of FUD on the fonts list |
| 03:34 | <annevk> | because font formats might not be safe they should be same-origin? give me a break |
| 03:41 | <othermaciej> | annevk: by "not safe" they mean cause security exploits like buffer overruns? |
| 03:41 | <othermaciej> | same-origin is clearly not even remotely a defense |
| 03:41 | <othermaciej> | against that threat model |
| 03:42 | <annevk> | read and weep: http://lists.w3.org/Archives/Public/public-webfonts-wg/2010Apr/0054.html |
| 03:42 | <othermaciej> | I wonder if Apple ended up joining the Fonts WG or not |
| 03:42 | <othermaciej> | if we did then I guess maybe I should join the list |
| 03:44 | <annevk> | you haven't |
| 03:44 | <annevk> | afaict |
| 03:45 | <annevk> | see also: http://www.w3.org/2000/09/dbwg/details?group=44556 (W3C Member-only) |
| 03:46 | <othermaciej> | it is true that fonts are a potential exploit vector, beyond what (non-SVG) images can do |
| 03:46 | <annevk> | I'm not at all disputing that |
| 03:46 | <othermaciej> | but same-origin is not an effective defense |
| 03:50 | <annevk> | and the claim elsewhere in the thread that fonts are more special than scripts is also utter crap |
| 03:50 | <annevk> | in http://lists.w3.org/Archives/Public/public-webfonts-wg/2010Apr/0050.html |
| 03:57 | <othermaciej> | it seems like a plausible argument that fonts are more widely reusable and less contextual |
| 03:58 | <othermaciej> | but that's not really a security argument |
| 04:00 | <othermaciej> | people also seem to have confused beliefs about what a same-origin restriction would do |
| 04:01 | <annevk> | oh, I didn't notice that |
| 04:01 | <othermaciej> | e.g. http://lists.w3.org/Archives/Public/public-webfonts-wg/2010Apr/0044.html |
| 04:01 | <othermaciej> | "My clients are interested in serving their custom corporate fonts in a way that does not expose them to use by other parties" |
| 04:03 | <annevk> | oh right, I was thinking he meant some kind of intranet situation |
| 04:03 | <annevk> | maybe I should do a reply after all |
| 04:04 | <othermaciej> | I don't think he means preventing external sites from using intranet-hosted fonts when viewed inside the firewall |
| 04:04 | <othermaciej> | I think he means preventing others from using a font that's published on a public site and protected by a same-origin restriction |
| 04:05 | <annevk> | yeah, so that's utter crap |
| 04:05 | <annevk> | wget! |
| 04:06 | <roc> | I think there is a significant difference between using wget to copy a font from someone else's server to your server, and serving it from your server |
| 04:06 | <roc> | vs putting a reference to someone else's server in your style sheet |
| 04:07 | <annevk> | othermaciej, btw, since you implemented <section> and all you should probably also implement the h1 styling rules from http://www.whatwg.org/specs/web-apps/current-work/complete.html#fonts-and-colors |
| 04:07 | <annevk> | othermaciej, they're after the initial block |
| 04:08 | <annevk> | roc, that's not quite the scenario he is sketching though |
| 04:08 | <annevk> | roc, also, I don't think anyone is against having a header for font files that says From-Origin: same or some such |
| 04:10 | <roc> | what would that do? |
| 04:11 | <annevk> | if that header is present and the UA will only use the resource if its same-origin with the Document it will be used for |
| 04:12 | <othermaciej> | annevk: yeah, we'll do that at some point |
| 04:12 | <othermaciej> | annevk: I guess either with a combinatorial explosion of rules, or we'll implement -webkit-any() |
| 04:12 | <annevk> | othermaciej, I'm afraid that if you wait sites will start depending on the default style rules |
| 04:13 | <othermaciej> | we also didn't implement the default margins yet |
| 04:13 | <othermaciej> | annevk: well, sites can use <section> just fine in browsers with 0 support |
| 04:14 | <annevk> | yeah, all the more reason to get the default style rules out there, but I see your point |
| 04:14 | <othermaciej> | our native support effectively just adds display: block, makes the interface HTMLElement instead of HTMLUnknownElement, and adds implicit <p> closing (and similar parsing effects) |
| 04:14 | <othermaciej> | oh, it looks like Hixie took away the default margins for sectioning elements |
| 04:15 | <roc> | annevk: I believe the number of sites that want same-origin checks will greatly exceed the number of sites that want to serve fonts cross-origin |
| 04:16 | <othermaciej> | anyway, I'll have to ask hyatt if adding those rules for h1 would cause any perf issues |
| 04:16 | <othermaciej> | I vaguely recall that rules with lots of indirect ancestor combinators can be slow |
| 04:16 | <othermaciej> | so it might be worthwhile to add a :-webkit-section-level() pseudo or the like |
| 04:17 | <annevk> | roc, I believe most sites won't care, much like most don't care about other resources |
| 04:17 | <roc> | I believe that too |
| 04:18 | <annevk> | creating special solutions for fonts seems backwards to me |
| 04:18 | <roc> | well, it depends on how widespread commercial fonts become on the Web and how much Web authors care about their license agreements |
| 04:19 | <roc> | imposing a same-origin restriction for fonts really isn't all that special. It's not like we invented same-origin restrictions. |
| 04:20 | <annevk> | well I think it is special |
| 04:21 | <roc> | ok |
| 04:21 | <annevk> | so far we only used same-origin restrictions if certain conditions apply |
| 04:21 | <annevk> | they don't apply to fonts |
| 06:28 | <JonathanNeal_> | heyo! |
| 06:30 | <alt-dot-net-geek> | hello |
| 08:04 | <annevk> | hmm, so fetch does not disable Referer if the origin passed in is a unique identifier |
| 08:04 | <annevk> | maybe I should ask for a Referer flag or something |
| 08:12 | <othermaciej> | it seems like Fetch should disable the referer in that case |
| 08:12 | <othermaciej> | er |
| 08:12 | <othermaciej> | wrong capitalization but you get the idea |
| 08:12 | <othermaciej> | it would never be right to send a unique origin referer |
| 08:13 | <annevk> | well, the Document in question could still have an origin |
| 08:14 | <annevk> | and an associated URL |
| 08:14 | <othermaciej> | annevk: so Tyler says anonymous requests (or whatever we call them) should discard Set-Cookie headers in the response |
| 08:14 | <othermaciej> | annevk: are you going to do that? I don't think it is an automatic side effect of unique identifier origin |
| 08:14 | <othermaciej> | he also mentioned "It would also need to prohibit the reuse of a TLS |
| 08:14 | <othermaciej> | client authenticated connection or NTLM authenticated connection. It |
| 08:14 | <othermaciej> | would also need to prevent use of cache entries populated by |
| 08:14 | <othermaciej> | non-uniform requests." |
| 08:14 | <annevk> | what I just checked in discards setting cookies when the credentials flag is false |
| 08:15 | <othermaciej> | I see he also mentioned Referer so maybe you are looking at his issues |
| 08:15 | <annevk> | so it's affected by withCredentials not by unique origins |
| 08:15 | <othermaciej> | I need to re-review the drafts for possible subset violations |
| 08:15 | <annevk> | but for the "new API" it comes down to the same thing |
| 08:15 | <othermaciej> | yeah it should be withCredentials based |
| 08:15 | <othermaciej> | that makes sense |
| 08:15 | <annevk> | (as the new API has credentials set to false and a unique origin) |
| 08:16 | <annevk> | I'm not quite sure what I should say exactly about TLS NTML and cache entries |
| 08:16 | <othermaciej> | did you settle on a name yet? |
| 08:16 | <othermaciej> | I apologize for splashing so much paint on the bikeshed |
| 08:16 | <annevk> | not really |
| 08:16 | <annevk> | guess I should read that thread with suggestions one more time |
| 08:16 | <othermaciej> | I think TLS and NTLM should be effects of withCredentials |
| 08:17 | <othermaciej> | and would be covered by a suitable definition of user credential |
| 08:17 | <othermaciej> | NTLM does per-connection auth, you can't really share a connection that's been NTLM-authenticated for no-credential requests |
| 08:18 | <annevk> | we could do ContextFreeRequest I suppose which inherits from XMLHttpRequest and is therefore on the same prototype chain and all |
| 08:18 | <othermaciej> | it's a bit of a computer-sciencey name |
| 08:18 | <othermaciej> | though not obviously worse than the others |
| 08:19 | <annevk> | so maybe I have TLS/NTML already: http://dev.w3.org/2006/waf/access-control/#credentials |
| 08:19 | <annevk> | though I should rename credentials to user credentials |
| 08:19 | <annevk> | UniformRequest makes no sense to me |
| 08:19 | <annevk> | GuestRequest is pretty good |
| 08:20 | <annevk> | incognito is too hard to spell |
| 08:22 | <hsivonen> | Is discarding cookies something that Caja can control if UMP is implemented on top of XHR? |
| 08:22 | <annevk> | no |
| 08:22 | <annevk> | XHR is cookie-ignorant |
| 08:22 | <annevk> | API-wise that is |
| 08:22 | <hsivonen> | does Set-Cookie in an XHR response set cookie in the browser HTTP stack? |
| 08:22 | <annevk> | yes |
| 08:23 | <annevk> | or just GuestXMLHttpRequest |
| 08:23 | <annevk> | I'm not too interested in this rhetorical context Mark went on about |
| 08:25 | <annevk> | so remaining issues |
| 08:25 | <annevk> | 1) Referer handling -- should maybe be fixed in HTML5 |
| 08:25 | <othermaciej> | annevk: it would be nice if the name was something that implies "dropping privileges" |
| 08:25 | <othermaciej> | hsivonen: Set-Cookie is done in the http stack but I believe http stacks expose enough that the browser could inhibit it per-request |
| 08:25 | <annevk> | 2) some kind of replacement for new XMLHttpRequest(true) |
| 08:26 | <othermaciej> | annevk: I could have sworn HTML5 handled Referer |
| 08:26 | <annevk> | 3) maybe some twiddling with the definition of credentials and at least renaming it |
| 08:26 | <annevk> | othermaciej, it does, but it does not allow for disabling it |
| 08:27 | <annevk> | othermaciej, what we have here is a Document object with a known origin and an API that wants a unique origin; all HTML5 allows for here is including the Referer and setting Origin to null |
| 08:27 | <othermaciej> | annevk: "If the origin of the appropriate Document is not a scheme/host/port tuple, then the Referer (sic) header must be omitted, regardless of its value." |
| 08:27 | <annevk> | (i.e. context-free requests) |
| 08:27 | <othermaciej> | http://dev.w3.org/html5/spec/Overview.html#fetch Step 1 |
| 08:28 | <annevk> | othermaciej, right, but the origin of _Document_ is a tuple, it's just that the XMLHttpRequest origin is not |
| 08:29 | <othermaciej> | annevk: ah - Referer is obtained from the Document's origin, not the passed-in origin |
| 08:29 | <othermaciej> | annevk: that is weird |
| 08:29 | <othermaciej> | annevk: perhaps Referer should also be dropped when the passed-in origin does not match the document's origin |
| 08:30 | <othermaciej> | I don't know what cases besides this new XHR feature use an explicit origin different from the document origin |
| 08:32 | <othermaciej> | I wish it was easy to find all references to the fetch algorithm |
| 08:32 | <annevk> | yeah, something like that might make sense |
| 08:32 | <othermaciej> | it's kind of hard to find which ones, if any, pass in an explicit origin |
| 08:32 | <annevk> | in the WHATWG copy you can click on "fetch" |
| 08:32 | <othermaciej> | I suspect nothing in HTML5 proper does |
| 08:32 | <annevk> | not sure if the W3C has the backreferences script |
| 08:32 | <othermaciej> | sold! |
| 08:32 | <othermaciej> | I only use the w3c copy because it's higher in the google rankings and I'm too lazy to type either URL |
| 08:32 | <othermaciej> | or even autocomplete |
| 08:34 | <othermaciej> | EventSource explicitly passes an origin, but the one it passes explicitly is the default one that would be used for an API |
| 08:34 | <annevk> | browsers should make the faster path to sites more convenient |
| 08:34 | <othermaciej> | likewise for importScripts |
| 08:34 | <othermaciej> | if I visit the spec enough times in a week it ends up in my Safari Top Sites |
| 08:34 | <othermaciej> | but I don't read it quite enough for it to stay there |
| 08:34 | <annevk> | what I think is actually the case here is that the origin argument is just used to determine whether the Origin header should be set to null or not |
| 08:34 | <othermaciej> | though I guess I could pin it if I cared to |
| 08:35 | <othermaciej> | the origin argument isn't actually passed in the Origin header? |
| 08:35 | <othermaciej> | no, that's not right "For the purposes of the Origin header, if the fetching algorithm was explicitly initiated from an origin, then the origin that initiated the HTTP request is origin." |
| 08:36 | <othermaciej> | I see - when the origin is *not* passed in, then you do "Origin: null" |
| 08:36 | <othermaciej> | kinda wacky |
| 08:37 | <annevk> | it's not very convenient either because you need to know beforehand whether to pass it or not based on whether it is a tuple or a unique identifier |
| 08:37 | <othermaciej> | if it's a unique identifier then I think it will produce "Origin: null" anyway |
| 08:37 | <othermaciej> | I guess the other places that hide the origin don't want to hide the referer |
| 08:38 | <annevk> | guess I should check up on that in the Origin draft |
| 08:39 | <annevk> | ah yes, you're right |
| 08:40 | <annevk> | the Origin draft deals with serialization |
| 08:40 | <annevk> | http://tools.ietf.org/html/draft-abarth-origin for people following at home |
| 08:44 | <annevk> | (renamed credentials to user credentials) |
| 08:47 | <hsivonen> | has any browser implemented the Origin header yet? |
| 08:47 | <annevk> | Gecko and WebKit I think |
| 08:48 | <annevk> | oh, and IE for XDomainRequest |
| 08:48 | <hsivonen> | annevk: for all requests or just for XHR? |
| 08:51 | <annevk> | for all I believe |
| 08:51 | <othermaciej> | I think WebKit implements it for all |
| 08:51 | <othermaciej> | but I am not sure |
| 08:53 | <hsivonen> | hmm. I can't easily test, because the Gecko builds I use are always ahead of extensions |
| 08:53 | <hsivonen> | the good thing about Opera's kitchen sink approach is that the kitchen sink is built with the rest of the app |
| 08:54 | <hsivonen> | ok. I managed to override Live HTTP Headers extension compat |
| 08:54 | <hsivonen> | it seems Gecko doesn't send Origin for all requests |
| 08:56 | <hsivonen> | even the GET to whatwg revision.dat doesn't have Origin according to Live HTTP Headers |
| 08:57 | hsivonen | sees "However, a centralized repository for mappings is an obviously desirable thing." in http://people.xiph.org/~xiphmont/lj-pseudocut/o-response-1.html |
| 08:58 | hsivonen | wonders if the TAG would agree |
| 08:59 | <othermaciej> | yeah, why don't they just embed URIs in the code stream, that map to prefixes, which are then embedded in each packet to identify the mapping |
| 08:59 | <othermaciej> | s/code stream/codec stream/ |
| 09:00 | <othermaciej> | that's the more natural way to do it |
| 09:06 | <annevk> | filed http://www.w3.org/Bugs/Public/show_bug.cgi?id=9603 and http://www.w3.org/Bugs/Public/show_bug.cgi?id=9604 |
| 09:07 | <gsnedders> | Dashiva: Hey, that still gives me a year to get married! |
| 09:07 | <annevk> | I handled setting cookies in a generic section for now, but it might make more sense to have it as part of fetching, since it is handled there already |
| 09:07 | <gsnedders> | Dashiva: And seeming I'm meant to be engaged to a Swedish girl by the end of August, that's plenty of time |
| 09:07 | <gsnedders> | *just under two years to get married |
| 09:40 | <jgraham> | gsnedders: GLWT |
| 09:57 | gsnedders | feels bad having to look up what GLWT means |
| 09:58 | <jgraham> | gsnedders: It's short for "all the best looking boys are taken, all the best looking girls are staing inside" |
| 09:59 | <jgraham> | *staying |
| 10:04 | <gsnedders> | So can I get a good looking, but not best looking girl/boy? |
| 10:09 | <annevk> | or maybe we should just stick to AnonXMLHttpRequest() |
| 10:09 | <annevk> | it wasn't a bad name |
| 10:09 | annevk | sighs |
| 10:12 | <roc> | it sounded good to me |
| 10:12 | <roc> | "Uniform" is a total head-scratcher |
| 10:15 | jgraham | wonders what the suggestions have been |
| 10:15 | <jgraham> | "GuestXML..." sounds very dubious to me |
| 10:16 | <jgraham> | I can't think of anywhere else on the Web where "Guest" is used to mean "without sending some data" |
| 10:16 | <jgraham> | And it's not obvious that that's what it does mean from the name |
| 10:17 | <annevk> | outside the web it usually means something like what it means here |
| 10:17 | <annevk> | guest accounts |
| 10:17 | <jgraham> | Yeah, I guessed that's where it came from |
| 10:17 | <jgraham> | But I still don't think it is obvious |
| 10:17 | <annevk> | would be great if there was any web platform precedent |
| 10:18 | <hsivonen> | XMLHttpRequestNoLogin |
| 10:19 | <annevk> | that would get the same objections as Anon |
| 10:19 | <Philip`> | XMLHttpRequest2 |
| 10:19 | hsivonen | is unaware of the objection |
| 10:19 | <annevk> | see this subthread: http://lists.w3.org/Archives/Public/public-webapps/2010AprJun/thread.html#msg262 |
| 10:19 | <annevk> | somewhere around there |
| 10:20 | <annevk> | Philip`, no |
| 10:20 | jgraham | is unaware of the whole thread other than what has been discussed here |
| 10:20 | <annevk> | alright, going for food now |
| 10:20 | <annevk> | will check logs later |
| 10:20 | <hsivonen> | annevk: anyway, I agreewith zcorpan's reasoning for having a suffix instead of a prefix |
| 10:22 | jgraham | has exactly no idea what semantic "Uniform" is supposed to convey |
| 10:23 | <Dashiva> | The same as URL, that is, nothing useful |
| 10:42 | <gsnedders> | So I found out yesterday that I have just over a week to decide about uni, which goes against what the guide said about when the deadline to decide was |
| 10:49 | <Philip`> | I thought you'd already spent about two years trying to decide, so another few weeks or months wouldn't make any difference |
| 10:53 | <gsnedders> | I thought three |
| 10:54 | <Philip`> | Three is about two |
| 10:57 | <gsnedders> | jgraham just bullied me into accepting it. |
| 10:57 | <gsnedders> | Now he rolling his eyes |
| 10:57 | <gsnedders> | And saying I speak like a foreigner |
| 10:57 | <gsnedders> | I must be Swedish already |
| 10:57 | <gsnedders> | No need to get married and stay here now |
| 11:59 | <hsivonen> | hmm. it sure is confusing that in the case of multiple <base> elements, the first one with href takes effect for href and the first one with target takes effect for target |
| 12:05 | <zcorpan> | hsivonen: is that implemented in gecko? |
| 12:07 | <hsivonen> | zcorpan: as of this week, yes |
| 12:32 | <mut> | hey, on canvas, how can i display a degrees symbol in text... like ° |
| 12:34 | <gsnedders> | ctx.fillText("°", 0, 0)? |
| 12:34 | <mut> | hmm its not showing |
| 12:34 | <mut> | heh just shows a ? |
| 12:34 | <mut> | in a blackbox |
| 12:34 | <gsnedders> | ctx.fillText("\u00B0", 20, 20) if you don't want to actually have non-ASCII chars |
| 12:34 | <roc> | maybe your font doesn't have it? |
| 12:34 | <mut> | ah |
| 12:35 | <roc> | what browser? |
| 12:35 | <mut> | ;) thanks |
| 12:35 | <mut> | err firefox |
| 12:35 | <mut> | chroime |
| 12:35 | <roc> | hmm |
| 12:35 | <gsnedders> | Yeah, that's the other likely thing |
| 12:35 | <mut> | and ie |
| 12:35 | <roc> | wait |
| 12:35 | <roc> | IE? |
| 12:35 | <mut> | yea with plugin |
| 12:35 | <mut> | sorry |
| 12:35 | <roc> | OK, no idea |
| 12:35 | <roc> | Firefox should display something if *any* font in your system supports that character |
| 12:35 | <roc> | which seems ... likely |
| 12:35 | <mut> | sweet that works now |
| 12:35 | <mut> | in ie too ;) |
| 12:51 | <jgraham> | hsivonen: Party time! |
| 12:51 | <gsnedders> | hsivonen: Congrats on getting it enabled by default at last |
| 12:51 | <hsivonen> | jgraham: yeah. and time to update the validotor, too. :-) |
| 12:51 | <hsivonen> | gsnedders: thanks |
| 12:54 | <gsnedders> | I guess the real question is whether you now get a load more bugs reported. |
| 12:54 | <hsivonen> | yes. interesting times. |
| 12:58 | <roc> | I'm glad our parser is deterministic now |
| 12:58 | <roc> | at least I hope it is! |
| 13:03 | <hsivonen> | roc: the tree shape should be deterministic now but the network loads started as the side effect of the parse aren't |
| 13:29 | <zcorpan> | hsivonen: data:text/html;charset=utf-8,x%c2 |
| 13:30 | <zcorpan> | hsivonen: should give an U+FFFD, right? doesn't with html5.enable (haven't tested without) |
| 13:30 | <hsivonen> | zcorpan: yes, it should. thanks |
| 15:23 | Dashiva | wonders if the sarcasm tag will survive last call |
| 15:25 | <gsnedders> | Dashiva: Which sarcasm tag? |
| 15:25 | <Dashiva> | The one in HTML5 |
| 15:25 | <gsnedders> | Ah, the end sarcasm tag ;P |
| 15:27 | <Dashiva> | It's a shame you need <sarcasm> as well to avoid the parse error |
| 15:27 | <jgraham> | Dashiva: What is this last call of which you speak |
| 15:28 | <jgraham> | ? |
| 15:28 | <Dashiva> | It's a mythical golden age prophecized to occur in the near future |
| 15:29 | <Dashiva> | When all bugs and issues will be vanquished, and leading to the implementations rising to claim their heritage |
| 15:30 | <jgraham> | When have prophets even been right? |
| 15:30 | <Philip`> | jgraham: Whenever they've made a sufficient number of sufficiently vague predictions that some are bound to be interpreted after the fact as right? |
| 15:31 | <CoJaBo-Aztec> | the economy sucks. Not much prophets now :P |
| 15:32 | <cardona507> | good morning :) |
| 15:32 | <jgraham> | Philip`: This "last call" thing sounds altogether too specific in that case |
| 15:32 | <Dashiva> | No, no |
| 15:32 | <Dashiva> | You assume a concrete definition of bugs and issues in that case |
| 15:35 | <gsnedders> | Sitting cross-legged on my chair always ends up with on my legs being numb |
| 15:35 | <gsnedders> | That makes standing up challenging |
| 15:35 | <hsivonen> | well, the HTML5 parsing fun didn't last. http://hg.mozilla.org/mozilla-central/rev/ccb50d524490 |
| 15:37 | <CoJaBo-Aztec> | randomorange? |
| 15:37 | <jgraham> | Orange == failed tests |
| 15:38 | <jgraham> | random orange == random failed tests |
| 15:38 | <jgraham> | == bad |
| 15:38 | <jgraham> | (I think) |
| 15:39 | <davidb> | yes random == hard |
| 15:39 | <gsnedders> | Random tests are my favourite. |
| 15:40 | <davidb> | what about random failures? :) |
| 15:40 | <jgraham> | The "I think wasn't about "== bad" |
| 15:40 | <jgraham> | s/k/k"/ |
| 15:40 | <gsnedders> | davidb: They're just a subset of them |
| 15:40 | <davidb> | jgraham all your statements looked correct |
| 15:43 | <jgraham> | How does Mozilla deal with tests that currently fail? Are they just not run or? |
| 15:43 | <davidb> | Everything is a judgement call, but generally we back out anything that regresses tests |
| 15:44 | <jgraham> | Right but I mean if you, say, add the CSS testsuite but know that you don't support all of the functionality yet |
| 15:44 | <jgraham> | So some tests fail |
| 15:45 | jgraham | remembers the expected fail state |
| 15:45 | <jgraham> | but if an expected fail passes does it have to be manually updated to expected pass or? |
| 15:50 | <davidb> | jgraham: yeah there are expected fail type of options, as well as per test things like mochitest "todo" etc. lots of choices, but generally requires manually updating when they pass |
| 15:50 | <davidb> | sorry for lag, i'm multi-chatting/tasking |
| 15:50 | <davidb> | as are we all i know :) |
| 15:52 | <hsivonen> | jgraham: upstream tests don't get run before someone sees them as green locally and pushes them to mozilla-central |
| 15:53 | <jgraham> | davidb: np. Thanks for sharing :) |
| 18:27 | <jgraham> | hsivonen: Ah. That is quite different from our approach |
| 19:03 | <davidb> | jgraham: what is your approach? |
| 19:18 | <JonathanNeal> | Hey everybody. |
| 19:46 | <jgraham> | davidb: Basically we record the result of every test in every build and look for changes between builds |
| 19:47 | <davidb> | i see |
| 19:47 | <davidb> | regressions are what matter |
| 19:48 | <jgraham> | Well yes, they get particular attention. |
| 20:33 | <tonyg-cr> | Does anyone happen to know if first posts to public-webapps⊙wo are moderated and if so, who approves them? |
| 20:35 | <jwalden> | I don't believe they're moderated, but my knowledge is about a year old (I think) |
| 21:13 | <webben> | maybe interesting data point wrt OCR of images with missing @alt: the WebVisum extension (current status, maybe iffy, some of their links are down) for firefox actually did this: http://preview.tinyurl.com/27e8ubo |
