02:50
<MikeSmith>
Hixie: here now, if you can remember what you had pinged me about
03:10
<Hixie>
MikeSmith: could you create an "HTML - <img>" component in Bugzilla, with simonp⊙oc as the default assignee, auto cc ian⊙hc?
03:11
<MikeSmith>
yup
03:11
<Hixie>
thanks!
03:16
<MikeSmith>
Hixie: OK done
03:16
<MikeSmith>
Hixie: I i
03:16
<Hixie>
thanks!
03:16
<MikeSmith>
I like the new tilted notes/warnings thing
03:17
<MikeSmith>
when did you change that?
03:17
<Hixie>
few days ago
03:17
<Hixie>
glad you like it :-)
06:53
<anvaka>
i'm curious. If html was invented today, how would it look like?
06:55
<anvaka>
would it be still xml-like? or maybe json? or maybe something totally different?
07:19
<MikeSmith>
tyoshino____: should we just replace the content at https://dvcs.w3.org/hg/streams-api/ with a pointer to https://github.com/whatwg/streams ?
07:19
<tyoshino____>
it includes a link to WHATWG one
07:20
<tyoshino____>
though it's not easy to find
07:27
<tyoshino____>
Status of This Document
07:27
<tyoshino____>
...
07:27
<tyoshino____>
The core primitives of the Streams API is now being developped at WHATWG GitHub repository. Please join us to finalize the core primitives. Once it's done, it's planned to be ported to here and extended to meet requirements specific to browser environment.
07:34
<MikeSmith>
tyoshino____: OK
09:55
<tobie>
jungkees: what's the policy for merging changes in the SW repo?
09:59
<jungkees>
For the web spec, you can make PR then Alex any myself will merge that
10:01
<jungkees>
tobie: for other part of the resources, there are quite a few contributors who have the rights I guess
10:01
<tobie>
jungkees: I have commit rights--just wanted to do the right thing. :)
10:02
<jungkees>
tobie: I see :-)
10:02
<tobie>
shouldn't the canonical spec be on the gh-pages branch, btw?
10:02
<tobie>
And not in the master branch?
10:02
<jungkees>
It already is
10:03
<tobie>
oh, is there a cron-job somewhere automatically copying things over?
10:03
<jungkees>
tobie: but basically we are rebasing gh-pages based off of master
10:03
<jungkees>
tobie: so please make changes in the master branch
10:04
<jungkees>
tobie: I don't think there is any automatic stuff
10:05
<tobie>
jungkees: setting up gh-pages that way would have the advantage that when a pull request is merged, the published spec is automatically up to date.
10:06
<tobie>
Right now it's not.
10:06
<jungkees>
tobie: I don't the how to; do you have any pointers that I can refer to?
10:07
<tobie>
I do: http://tobie.github.io/specs-on-github/
10:07
<jungkees>
tobie: thanks
10:08
<tobie>
jungkees: happy to make a PR for your consideration.
10:08
<jungkees>
tobie: thanks tobie!
10:10
<tobie>
jungkees: mind merging https://github.com/slightlyoff/ServiceWorker/pull/265 first?
10:11
<jungkees>
tobie: For web-spec-framework part of it, I would like for Alex to review
10:11
<jungkees>
tobie: thinks he's recovering from surgery now..
10:11
<tobie>
Oh. Wasn't aware.
10:31
<tobie>
jungkees: opened a ticket here: https://github.com/slightlyoff/ServiceWorker/issues/266
10:33
<jgraham>
MikeSmith: yt? Do you have a naming scheme for conformance checker tests? /conformance-checkers/html-aria/roles-properties-supported-inherited/roles-properties-supported-inherited-progressbar-aria-valuetext-Test-string-value.html
10:34
<jgraham>
is longer than 150 characters
10:34
<jgraham>
Which I'm trying to set as the limit for windows compat (sigh)
10:56
<jgraham>
Does anyone who isn't MikeSmith know how the conformance checker test naming scheme works?
11:03
<MikeSmith>
jgraham: ARIA test are imported. I didn't name them. but I can rename them if need. and rename any others. there's not any consistent naming conventions that must be followed there. my test harness doesn't care about the names.
11:06
<jgraham>
MikeSmith: OK. Could we maybe remove the duplication of having rols-properties-supported-inherited in both the directory name and the test title?
11:06
<jgraham>
*roles
11:06
<jgraham>
s/test title/file name/
11:11
<MikeSmith>
jgraham: sure. I'll change them today
11:17
<jgraham>
MikeSmith: https://critic.hoppipolla.co.uk/8c72c580?review=1516 is my attempt. Changes enough to meet the length limits, but not everything systematically
11:47
<MikeSmith>
jgraham: r+ed that part of the review
11:47
<MikeSmith>
btw it would be nice if critic indicated that files had been renamed, the way git log does and the github view does
11:48
<MikeSmith>
instead of just showing added and removed
11:50
<jgraham>
MikeSmith: Yeah :(
11:54
MikeSmith
finds http://lint.travis-ci.org/
11:56
<sangwhan>
MikeSmith: have you tried the shortcut "m"?
11:58
<sangwhan>
MikeSmith: that will do a second pass scan for renamed/moved files
11:59
<sangwhan>
on directories with lots of very similar files it might trip though
12:18
<MikeSmith>
sangwhan: oh cool
12:18
<MikeSmith>
hadn't tried that
12:19
<MikeSmith>
yeah in this case though it does seem to trip up
13:38
<zewt>
git seems do renames heuristically too, for some reason
13:39
<zewt>
whenever i add a new file that looks like an old one and delete the old one, it shows up in git status (while i'm staging) as a move and change instead of an add and delete (which is really annoying)
13:59
<odinho>
zewt: that's the way it figures out renames. It only tracks content, not metadata like that.
13:59
<odinho>
zewt: Do separate commits if you like.
14:00
<Ms2ger>
But git is superior to hg, because tools can always figure out file moves perfectly after the fact, without any metadata!
14:00
<odinho>
(renames are just delete+add in git anyway, no difference between it, -- the rename is just syntactical sugar to help show what is going on)
14:00
<zewt>
yeah, it's pretty lame
14:01
<jgraham>
Oddly enough git is superior to hg because it has a better cli
14:01
<odinho>
And better model.
14:01
<jgraham>
If you don't believe me, try using mercurial bookmarks for a a while
14:01
<odinho>
Since hg's is mindfuck :)
14:01
<Ms2ger>
Git has a better cli?
14:02
Ms2ger
is baffled
14:02
<jgraham>
Yeah, it does help that it's possible to actually understand how git works without having to read the source
14:02
<Ms2ger>
... You're joking, right?
14:02
<jgraham>
Ms2ger: I agree this is not the conventional wisdom
14:03
<zewt>
with git you just have to read the whole internet instead
14:03
<jgraham>
(about the cli, not about the model)
14:03
<odinho>
Is it not the conventional wisdom? I've tried and failed to understand hg for weeks and weeks when I had to use it.
14:03
<odinho>
It was horrible. -- Git on the other hand had some explainers of the model that was immediately obvious to me.
14:03
<mark06>
odinho: how does the diff looks like for a "renamed" file?
14:04
<Ms2ger>
odinho, I guess it's hard if you're used to git
14:04
<jgraham>
Conventional wisdom: git has a terrible cli but a simple model once you understand it. Reality: hg and git both have rough edges in their cli but only hg actively tries to prevent you from doing useful things. The hg model seems to be more complex than the git one and isn't really documented.
14:04
<Ms2ger>
And if you try to force git habits onto hg
14:05
<mark06>
zewt: looks like how exactly? you added a file with similar content next to old, deleted file or with same path?
14:05
<Ms2ger>
The reality is that git's is full of rough edges; in hg you have to opt in to them
14:05
<Ms2ger>
Example: git pull
14:05
<jgraham>
Ms2ger: FWIW the "hg community" now seems to believe that mq is a terrible idea and that git-style local branches are good.
14:05
<zewt>
mark06: the usual case where i hit it in practice is when i'm updating an external SDK, where i'll extract foo-1.2.3, delete the old foo-1.2.2 and update a symlink
14:06
<zewt>
i know what's happening, of course
14:06
<Ms2ger>
Well, the "hg community" isn't necessarily right, of course
14:06
<Ms2ger>
Anyway
14:06
<jgraham>
But in this case they have a point because mq defeats the whole point of a vcs unless you add a second vcs on top of the first
14:07
<zewt>
i guess if i was updating it in-place the magic-renaming thing would be better
14:07
<jgraham>
Also, re: no one understanding the model; Mozilla's Try server effectively becomes unusuable after too many people have pushed to it and no one seems to know why
14:08
<jgraham>
So the solution is "reset it every so often"
14:08
<zewt>
jgraham: future
14:08
<jgraham>
Which is laughable
14:08
<Ms2ger>
Sure
14:09
<Ms2ger>
Tryserver is an abuse of hg anyway
14:09
<jgraham>
Why?
14:09
<jgraham>
It seems like a totally legitimate use case
14:09
<mark06>
I find that web interface for hg very counterintuitive, and I saw a doc once on how to set up a mail-on-commit hook... I guess implementing an smtp server from scratch would be easier... I don't know git but it seems counterintuitive too... I use bzr, doesn't hurt me
14:10
<jgraham>
I don't know how try actually works but conceptually the model is just "create a remote branch"
14:10
<mark06>
zewt: if the symlink is versioned, isn't that just right? in fact it has changed
14:11
<zewt>
no, it shows "foo-1.2.3/main.c renamed to foo-1.2.4/main.c, ..."
14:12
<zewt>
afk
14:14
<mark06>
odd, I'd expect it not following the symlink
14:16
<jgraham>
(hg does have some nice features. And I think that changeset evolution could be a good thing, although it remains to be seen if it works well)
14:19
<sangwhan>
fwiw, hg has more sane defaults than git - git's defaults are very weird and do stuff completely unexpected. (see: git pull)
14:20
<odinho>
pull is created for pull-based workflows. But very few seem to use that. Except Linux kernel.
14:20
<sangwhan>
and git's windows support is very laughable
14:21
<odinho>
Git could work much better on Windows if someone who loved Windows would do an implementation though :)
14:23
<jgraham>
Aren't Microsoft using git these days (for some of their open source stuff)? Seems like it could get better if that's the case.
14:23
<davve>
jgraham: They're using libgit2 at least.
14:23
<odinho>
It is getting better. Because many things are stupid. Like msysgit shipping severely outdated openssh which has serious performance issues.
14:25
<sangwhan>
Not sure. Some features in git are implemented as shell script after all, and IIRC a significant amount of native has only a posix implementation. (Hence msys.)
14:25
sangwhan
might be incorrect, it's been several years since I checked
14:26
<mark06>
odinho: not also the heartbleed breach?
14:27
<sangwhan>
mark06: openssh shouldn't be affected by heartbleed?
14:27
<odinho>
mark06: not used as server, -- only client for downloading git repos etc.
14:27
<zewt>
mark06: symlink doesn't enter into it, it's just a file
14:27
<odinho>
plus what sangwhan said. that's openssl
14:28
<zewt>
git rm -rf foo-1.2.3; tar zxvf ~/foo-1.2.4.tar.gz; git add foo-1.2.4; rm foo; ln -s foo-1.2.4 foo; git add foo
14:28
<sangwhan>
on top of that msys/cygwin IO performance is abysmal when running the same code compared to any proper posix
14:28
<mark06>
I thought openssh used openssl for the backend
14:28
<sangwhan>
mark06: nope
14:29
<zewt>
odinho: still a problem (client has access to the user's private key), just less of one (would need a compromised server for it to matter)
14:29
<odinho>
zewt: okay. I don't know the area well at all :)
14:30
<zewt>
i don't really know details about the bug, either
14:31
<mark06>
ah I thought ssh was built on top of tls/ssl
14:32
<mark06>
zewt: so foo-1.2.3 is versioned, you delete it, add foo-1.2.4, and it thinks it's a rename?
14:33
<zewt>
yeah, since the files look the same
14:33
<zewt>
like odinho mentioned earlier
14:34
<zewt>
(the original discussion was about critic doing that; i was just observing that it's what git does)
14:34
<jgraham>
The original discussion was about critic *not* doing that
14:35
<jgraham>
Which it doesn't
14:35
<jgraham>
Unfortunately
14:35
<zewt>
11:53 < sangwhan> MikeSmith: have you tried the shortcut "m"?
14:35
<zewt>
11:55 < sangwhan> MikeSmith: that will do a second pass scan for renamed/moved files
14:35
<jgraham>
Yeah OK
14:35
<jgraham>
That is suboptimal in various ways, however
14:36
<jgraham>
(it isn't the default, you can't review files from that view, etc.)
14:36
<mark06>
also, msys (and msysGit) lacks timezone support completely, if you set $TZ for using date.exe correctly, MSVCRT applications get confused instead (IIRC git isn't really compile *for* msys)
14:37
<sangwhan>
zewt: there are some limitations with that feature, sadly
14:37
<sangwhan>
zewt: sort of works when most files have severely different content. i think you can adjust the threshold on the server side somehow if i remember correctly.
14:39
<mark06>
zewt: you say symlink doesn't enter into it, but your code snippet shows it's indeed being/getting versioned... why you add foo if foo-1.2.4 is already added?
14:41
<zewt>
i've never tried to use git in windows, i just use a linux shell with a cifs mount if i want a git wc to live in windows
14:41
<zewt>
mark06: so the build process knows where to find it
14:41
<zewt>
the symlink has nothing to do with the rename problem, it was just explaining why i'm adding and removing like that
15:00
<mark06>
it actually make it more confusing to me understand... but anyway... you explicitly "git rm" and "git add"... it's really annoying it insisting to mean a rename
15:02
<mark06>
but as someone said, there isn't a concept of rename and I guess it feels free to be "clever", expecting you to cheer on it
15:04
<mark06>
but my bet is it just groks it from the symlink...
15:23
<zewt>
(no, it uses the similarity of the files)
15:23
<mark06>
anyone knows about status of web speech api in firefox? it indeed has a speechSynthesis but looks useless... (e.g. no SpeechSynthesisUtterance or docs...)
15:25
<mark06>
that wouldn't hurt, would it? you're trying to cheat git but it is smarter haha...
15:26
<mark06>
initially, I wouldn't bother since they're actually the same files... but updated to new version...
16:00
<dglazkov>
good morning, Whatwg!
16:01
<IZh>
Good evening.
16:34
<eligrey>
i'm not very familiar with wiki syntax
16:35
<eligrey>
how do i rename http://wiki.whatwg.org/wiki/NavigatorCores to NavigatorHWConcurrency
16:35
<eligrey>
or at least redirect to that
16:35
<eligrey>
(from the current page, that is)
16:37
<Domenic_>
WebIDL question (seriously asking): why not use `partial interface Navigator`? I guess you'd have to duplicate between `partial interface Navigator` and partial interface WorkerNavigator`?
16:37
<Ms2ger>
Yes
16:37
<Domenic_>
OK, cool.
16:37
<Ms2ger>
eligrey, done
16:37
<eligrey>
thanks Ms2ger
16:37
<eligrey>
also is my Navigator implements NavigatorCPU stuff necessary in the idl?
16:38
<eligrey>
i wasn't sure but a chrome eng said it should include that
22:49
<caitp>
anyone want to look at merging https://github.com/w3c/web-platform-tests/pull/976? without that, all/most of the XHR worker tests are broken, pretty much
23:59
<Hixie>
MikeSmith: http://www.whatwg.org/specs/web-apps/current-work/multipage/text-level-semantics.html#the-time-element
23:59
<Hixie>
MikeSmith: enjoy the new styles :-P