03:09
<Domenic>
Wait so what is the relationship between composed events and event retargeting
03:10
<Domenic>
composed = false means the event doesn't make it past the shadow root boundary?
06:48
<MikeSmith>
anybody awake who has written web-platform-test tests and ever run into the error “Cannot read property 'has_extended_attribute' of undefined” when running idlharness.js tests on an interface
06:54
<annevk>
Domenic: correct
06:55
<annevk>
Domenic: well, the shadow root boundary that is the event's target root, if any
07:18
<Ms2ger>
zcorpan, welcome back :)
07:18
<zcorpan>
Ms2ger: thx
08:12
<MikeSmith>
det var länge sedan zcorpan
08:12
<zcorpan>
MikeSmith: o/
08:12
<MikeSmith>
glad you had a nice vacation but also glad you’re back :)
08:15
<zcorpan>
:-)
08:15
<zcorpan>
is there something i should deal with pronto?
08:17
<MikeSmith>
zcorpan: nope, though I reckon I will start recalling various things I’d wanted to ask you about
08:17
<MikeSmith>
zcorpan: dunno if you saw yet but I landed the ignore-template-subtree support in the HTML checker
08:18
<MikeSmith>
and it seems to work as expected
08:18
<MikeSmith>
it was a bit more complicated to implement than I had expected
08:19
<MikeSmith>
oh also you will find some interesting info at https://validator.w3.org/nu/stats.html if you scroll down
08:20
<MikeSmith>
zcorpan: ah one thing I wanted to ask you is please think about anything you think we should have the checker collect statistics on
08:20
<zcorpan>
MikeSmith: yeah i saw, nice
08:21
<MikeSmith>
e.g., I recently added a counter for <style> in body
08:22
<zcorpan>
Manually set character encoding 0 haha
08:23
<zcorpan>
hmm does v.nu still use us-ascii as default for text/xml?
08:23
<hsivonen>
zcorpan: possible
08:23
<botie>
hsivonen, at 2016-08-05 06:42 UTC, Ms2ger said: it :)
08:23
<zcorpan>
where is mozillaquestquest.com when you need it
08:24
<hsivonen>
:-)
08:24
<Ms2ger>
Thank you botie
08:24
<hsivonen>
fun fact: you can show the encoding UI even in the HTML5 UI using a magic URL: https://html5.validator.nu/?charset
08:24
<MikeSmith>
oh
08:26
annevk
runs git blame source
08:26
<annevk>
Maybe I should go make some tea
08:27
<MikeSmith>
hah
08:27
<MikeSmith>
for the HTML spec I find I have a lot more success just grepping through git log output than using blame
08:29
<MikeSmith>
because often the first blame doesn’t tell me what I want to know and then I have to recurse
08:30
<zcorpan>
MikeSmith: i recall hsivonen collected statistics of most common errors back in the day
08:30
<MikeSmith>
oh yeah?
08:31
<MikeSmith>
I wonder now how
08:31
<hsivonen>
zcorpan: very long ago. I haven't had that logging running in the recent years.
08:31
<annevk>
MikeSmith: grep failed to find it, or I did not know what to search for… 😞
08:32
<MikeSmith>
ah
08:32
<MikeSmith>
annevk: if you have to recurse you might find https://github.com/scottgonzalez/recursive-blame useful
08:33
<MikeSmith>
though it’s broke for me in my environment right now
08:34
<zcorpan>
MikeSmith: https://lists.w3.org/Archives/Public/public-html/2008Jan/0305.html
08:34
<hsivonen>
MikeSmith: https://github.com/validator/validator/blob/cbca37f2a633edcb512395e260e5e5887882b28f/src/nu/validator/messages/MessageEmitterAdapter.java#L745 is the logger than needs to be enabled in the log4j config
08:35
MikeSmith
looks
08:35
<MikeSmith>
oh wow
08:36
<MikeSmith>
hsivonen: will that just log all the messages? but not count them, right?
08:36
<MikeSmith>
zcorpan: thanks
08:37
<annevk>
zcorpan: gsnedders: do you happen to know where the parser handles duplicate attributes?
08:38
<hsivonen>
MikeSmith: counting was with post-processing using traditional shell tools
08:38
<MikeSmith>
hsivonen: though I realize it would not be hard to just post-process that with some sort+uniq or whatever
08:38
<MikeSmith>
hsivonen: yeah
08:38
<annevk>
Or maybe MikeSmith / hsivonen knows?
08:38
<annevk>
As far as I can tell the tokenizer hands duplicates to the tree builder, but they're not handled there
08:38
<annevk>
We have a patch to fix it there: https://github.com/whatwg/html/pull/1637
08:39
<annevk>
But I wonder whether I'm missing something
08:39
<zcorpan>
"When the user agent leaves the attribute name state (and before emitting the tag token, if appropriate), the complete attribute's name must be compared to the other attributes on the same token; if there is already an attribute on the token with the exact same name, then this is a parse error and the new attribute must be removed from the token."
08:40
<zcorpan>
we should inline that
08:40
<MikeSmith>
annevk: https://github.com/validator/htmlparser/blob/master/src/nu/validator/htmlparser/impl/Tokenizer.java#L1125
08:40
<annevk>
zcorpan: that might make sense
08:40
<hsivonen>
annevk: paragraph at the end of the Attribute name state: https://html.spec.whatwg.org/multipage/syntax.html#attribute-name-state
08:41
<MikeSmith>
code comment in the htmlparser source is the same zcorpan quoted
08:41
<annevk>
zcorpan: where is <html test > <html test=2 x> handled?
08:41
<MikeSmith>
and same hsivonen just cited
08:43
<annevk>
zcorpan: never mind, found it
08:47
<MikeSmith>
zcorpan: OK, I' experiment
08:47
<MikeSmith>
oofs
08:48
<MikeSmith>
zcorpan: OK, for the checker I’ll experiment with turning on logging of all error messages so we can have raw logs to get most-common-errors data from
08:49
<zcorpan>
MikeSmith: ok cool
08:49
<MikeSmith>
not sure how long I can keep it running for the W3C instances given how much they get hit, but I reckon even just 24 hours of logs should give us what we need
08:50
<MikeSmith>
altogether they process about a million document requests every day
08:51
<MikeSmith>
and I reckon each document on average must have a dozen or more errors
08:51
<MikeSmith>
documents with hundreds of errors are not uncommon
08:52
<MikeSmith>
anyway, will need to wait til later this week. Today I have to work on writing tests for the Payment Request API
08:54
<MikeSmith>
hey a somewhat interesting side effect of the language-detection stuff I added is that I also created a specific (n-gram) profile to detect "Lorem ipsum" text
08:55
<MikeSmith>
and it turns out that "Lorem ipsum" documents are one of the top-20 most common “languages” detected
08:56
<MikeSmith>
about 0.3% of all documents checked
08:58
<MikeSmith>
0.3% is more than, e.g., the number of Norwegian documents
09:23
<MikeSmith>
zcorpan: (and all) I notice that the HTML spec never normatively states what html@lang is
09:23
<MikeSmith>
instead it just says, “Authors are encouraged to specify a lang attribute on the root html element, giving the document's language.”
09:23
<MikeSmith>
should we make that normative
09:24
<zcorpan>
MikeSmith: it has normative text for the language of a node
09:25
<zcorpan>
https://html.spec.whatwg.org/multipage/dom.html#language
09:29
<MikeSmith>
ah OK
09:31
<MikeSmith>
so of node=HTMLHtmlElement, then it is the language for the entire document
10:29
<gsnedders>
annevk: the tokenizer does
10:29
<gsnedders>
annevk: gimme a min
10:29
<gsnedders>
annevk: oh, hsivonen already said
10:30
gsnedders
can't skim read when sleep depreived, obviously
11:09
<annevk>
Supporting the HTTP trailer is actually a whole lot of work
12:32
smaug____
wonders if one should follow wicg
12:32
<smaug____>
I guess no
14:00
<annevk>
Domenic: I noticed fetch() is using entry setting objects
14:14
<MikeSmith>
is there a way to run chromium third_party/WebKit/LayoutTests from the web?
14:14
<MikeSmith>
just whatever is in the trunk I mean
14:16
<caitp>
doubt it
14:17
<caitp>
since layout tests often depend on special APIs not exposed to the web, it seems like it would be hard to run them from the web
14:49
<MikeSmith>
caitp: ok, makes sense
20:41
<Domenic>
annevk: :(
21:38
<jsbell>
let the record show that boris zbarsky is an awesomely helpful person
21:40
<astearns>
in my understanding, the record is littered with those sentiments
22:48
<smaug____>
could someone explain why https://w3c.github.io/webappsec-mixed-content/#categorize-settings-object works
22:49
<smaug____>
it just goes through from document to subdocument and possibly its descendants
22:49
<smaug____>
but doesn't go through all the subdocuments of the top level doc
22:49
<smaug____>
mkwst: ^
22:52
<smaug____>
maybe I don't understand what "embedding document" means
22:52
<smaug____>
yes, I don't
22:52
<smaug____>
I don't know whether "embedding document" is ancestor or descendant
22:53
<smaug____>
Domenic: you might be awake
22:54
<smaug____>
or annevk
23:10
<jyasskin>
smaug____: It'll be the ancestor. Follow https://w3c.github.io/webappsec-mixed-content/#embedding-document and https://html.spec.whatwg.org/multipage/browsers.html#browsing-context-nested-through
23:10
<smaug____>
yup
23:11
<smaug____>
the language could be a bit simpler to read... or I should not read it at 2am
23:17
<jyasskin>
I think it could use "parent document" by analogy to "parent browsing context" (https://html.spec.whatwg.org/multipage/browsers.html#parent-browsing-context). It gives me a bit of pause that https://www.w3.org/TR/html51/browsers.html#sec-nested-browsing-contexts is quite different from
23:17
<jyasskin>
https://html.spec.whatwg.org/multipage/browsers.html#nested-browsing-contexts.
23:56
<Domenic>
Yeah, webappssec linking to the wrong spec causes a lot of problems for people trying to understand the specs :(
23:56
<Domenic>
Would be better to follow CSSWG policy and link to the fork only for REC snapshots, and leave EDs linking to the HTML Standard.