01:23
<MikeSmith>
caitp: would be good to have your perspective as an implementor on https://github.com/whatwg/html/pull/401 which is still open and welcoming comments
01:23
<MikeSmith>
caitp: also https://github.com/whatwg/html/pull/373 which has already been merged but seems like you might want to be aware of
01:24
<MikeSmith>
not sure if you care so much about stuff at this level, so just FYI
01:25
<caitp>
of course I care! thanks for the info
01:25
<caitp>
I'll have to have a look at them at a time when I haven't recently had a glass of wine
01:55
<MikeSmith>
heh
02:03
<smaug____>
annevk: Domenic: I need some spec reading help
02:03
<smaug____>
https://html.spec.whatwg.org/multipage/embedded-content.html#img-environment-changes
02:03
<smaug____>
step 2
02:03
<smaug____>
how is one supposed to interpret that
02:05
<smaug____>
especially, can one read it like "if the img element ....is not in a Document ..., then abort this algorithm"
02:10
<smaug____>
or does "it is not in a Document" refer to img's parent?
02:16
<caitp>
it looks to me like that step enumerates reasons to stop doing anything, and that part in particular refers to the <img> node
02:17
<caitp>
but what do I know, someone else can have a go at interpretting it
02:18
<MikeSmith>
if it's ambiguous then it's a bug and merits smaug____ or somebody taking time to raise an issue for it, IMHO
02:19
<MikeSmith>
because if it's not clear to the both of you then it's not going to be any more clear to other implementors
02:19
<smaug____>
caitp: but does that "it is not in a Document" refer only to <img> elements which don't have srcset?
02:20
smaug____
files a bug
02:20
<caitp>
it's a run-on sentence, probably written in a hurry, can't be sure
02:22
<caitp>
you can never tell in those cases whether someone is intentionally referring to the original subject, or if they had meant to split it into multiple sentences, or maybe use semicolons instead of commas --- all you can do is ask the author
02:25
<smaug____>
https://github.com/whatwg/html/issues/414 it is
09:05
<zcorpan>
can we rename data-x="" to lt=""? bikeshed habit
09:07
<annevk>
zcorpan: possibly, if you hack the toolchain and stuff
09:12
<zcorpan>
and also switch from <span> to <a> at the same time
09:13
<annevk>
zcorpan: yeah, seems trickier since we use <a> too, but who knows, I haven't looked at wattsi much
09:13
<zcorpan>
not particularly tricky, just leave those with href alone
09:22
<zcorpan>
maybe as a first step i can make wattsi fatal error when it sees href-less <a> or an lt="" attribute, which can catch mistakes now and make sure people upgrade wattsi later when we do switch to <a> and lt=""
10:57
<zcorpan>
MikeSmith: hmm i get an error when trying to build wattsi, even without any changes
10:57
<zcorpan>
Linking ../bin/wattsi
10:57
<zcorpan>
ld: file not found: /usr/lib/crt1.o
10:57
<zcorpan>
An error occurred while linking
10:57
<zcorpan>
wattsi.pas(2283) Error: Error while linking
10:57
<zcorpan>
wattsi.pas(2283) Fatal: There were 1 errors compiling module, stopping
11:13
<zcorpan>
PRed anyway https://github.com/whatwg/wattsi/pull/5
11:34
<smaug____>
zcorpan: "use srcset"
11:34
smaug____
wonders what that means
11:34
<smaug____>
perhaps 415 will explain that
11:34
<zcorpan>
smaug____: it's <dfn>ed
11:35
<annevk>
smaug____: https://github.com/whatwg/html/pull/406 requires your review
11:35
<zcorpan>
smaug____: https://github.com/whatwg/html/pull/415/files#diff-36cd38f49b9afa08222c0dc9ebfe35ebR24258
11:36
smaug____
has no idea how to see review queue in github easily
11:36
<smaug____>
zcorpan: ok, I see
11:37
<annevk>
smaug____: https://github.com/pulls/assigned
11:38
<smaug____>
zcorpan: so your interpretation is that if img isn't in document, return early, right?
11:38
<smaug____>
good
11:38
<smaug____>
since that is what I was hoping to see
11:38
<smaug____>
(since it makes fixing a bug easier in Gecko)
11:38
<zcorpan>
smaug____: yes. the in a document check was there before picture existed, i believe
11:39
<smaug____>
but I don't actually understand why in-document affects to srcset handling
11:41
<smaug____>
one can't do something like: var im = new Image(); im.srcset = "..."; canvas.getContext("2d").drawImage(...)
11:41
<smaug____>
but with src it works
11:41
<zcorpan>
smaug____: no you can do that. it just won't load new images when the environment changes
11:42
<smaug____>
well, that is what I meant. You need to manually load each time
11:43
<smaug____>
is there some reason why is-in-document matters for srcset?
11:44
<zcorpan>
ok. i'm not sure why the in-document check is there. if there are use cases maybe we should drop that
11:44
<smaug____>
annevk: so, how do I r+ in github
11:44
<annevk>
smaug____: just write a comment saying LGTM or r+
11:45
<smaug____>
huh
11:47
<smaug____>
can I do that comment in the view where I'm looking at the patch?
11:49
<smaug____>
I guess I can do that...
11:50
<zcorpan>
comment in the "Conversation". if you comment on a commit the comment disappears if the PR is rebased and force-pushed
11:50
<smaug____>
odd tool
11:50
<zcorpan>
or i suppose it doesn't disappear, but it is no longer visible in the PR
11:51
<zcorpan>
reviews in github isn't awesome...
11:52
<smaug____>
or there isn't reviews
11:52
<smaug____>
there are just some random comments
11:52
<annevk>
you can leave feedback on individual lines
11:53
<annevk>
but there isn't really a r+ thing and such
11:53
<annevk>
I'll merge that PR
12:27
<smaug____>
does github issue tool have some kind of dependency tracker?
12:38
<Ms2ger>
smaug____, no, but you can mention #1234 and that will show up in #1234
12:57
<smaug____>
Ms2ger: but if you close then the bug which had #1234 in a comment, #1234 isn't notified about it, right?
12:58
<Ms2ger>
Nope
13:01
<zcorpan>
smaug____: so is https://github.com/whatwg/html/pull/415 clear enough?
13:05
<smaug____>
zcorpan: I think so
13:05
<smaug____>
(I still don't know why srcset handling depends on is-in-document, when normal image loading doesn't.)
13:06
<smaug____>
but I can live with it depending in is-in-document
13:06
<smaug____>
just a minor inconsistency
13:06
<zcorpan>
smaug____: can you file a new issue about that?
13:07
<MikeSmith>
zcorpan: I have never seen that ld: file not found: /usr/lib/crt1.o error before
13:18
<smaug____>
zcorpan: https://github.com/whatwg/html/issues/416
13:19
<zcorpan>
smaug____: thx
13:23
<MikeSmith>
zcorpan: about that error, if you're on OSX, but you don't have XCode installed, maybe that's why
13:28
<zcorpan>
oh right, XCode always stops working whenever it updates, i think
13:33
<jgraham>
You need to accept a license agreement after every update
13:33
<jgraham>
In another example of
13:33
<jgraham>
"apple hate you:
13:35
<zcorpan>
plus xcode-select --install every time to redownload and install command line tools
14:22
<zcorpan>
ok it builds now. but Fail() isn't fatal
14:30
<zcorpan>
MikeSmith: why doesn't html-build show wattsi error messages?
14:45
<Domenic>
zcorpan: is wattsi in your PATH?
14:46
<zcorpan>
Domenic: it runs fine (now), it just doesn't log error messages from Fail() when using html-build/build.sh
14:46
<zcorpan>
Domenic: but it does when i run wattsi directly
14:46
<zcorpan>
Domenic: and we have a bunch of errors that i suppose nobody notices
15:09
<Domenic>
zcorpan: that seems bad, we should fix that asap :(
15:09
<Domenic>
zcorpan: I just thought maybe it was going to the web service which didn't have your local changes
16:04
<MikeSmith>
I wonder why Simon is running wattsi directly rather than using build.sh
16:04
<MikeSmith>
as far as I understand it, wattsi is not intended to be run directly on the raw source file
16:05
<MikeSmith>
if that's what he means by "directly"
16:07
<MikeSmith>
if he's running it without the pre-processing that the build.sh script is doing, then I think it's expected that there would be errors
16:07
<MikeSmith>
anyway I'l wait til he's back around
16:08
<MikeSmith>
botie, inform zcorpan here now if you have time to chat about the build script
16:08
<botie>
will do
16:36
<MikeSmith>
time for me to get back to ironing out the build wrinkles I guess
16:37
MikeSmith
digs his Builder cap out of the toy box and puts it back on
16:39
<MikeSmith>
two months is too long to be away from it I guess
16:39
<MikeSmith>
it has missed me
16:39
<MikeSmith>
nice to feel wanted
16:43
<Domenic>
yaaaa <3
16:56
<annevk>
Domenic: can you get from a Realm to a settings object?
16:56
<Domenic>
annevk: after my changes yes!
16:56
<annevk>
Domenic: the settings object has an origin
16:56
<annevk>
Domenic: don't think we want to use incumbent
16:56
<Domenic>
wasn't sure
16:57
<Domenic>
"the settings object whose realm execution context's Realm component is _realm_".
16:57
<Domenic>
We could shorten that by defining "a settings object's Realm" like I did for "a settings object's global object"
16:57
<annevk>
the wording of these things makes the checks look bizarro
16:57
<annevk>
yeah, I guess I'll wait until bz gets back from vacation
16:58
<annevk>
and myself
16:58
<annevk>
and then tweak things a bit and start writing a patch for HTML
16:59
<Domenic>
Very exciting to get these kind of long-term confusion issues straightened out
17:10
<rits>
Domenic: when i made the wrap_width changed to 100, now it is showing the changes in the whole paragraph
17:17
<annevk>
rbyers: as a heads up, I'll be away until January 1, seems like this is mostly done, but not done enough to wrap up before the holidays
17:18
<annevk>
rits: if it's the paragraph you're making changes too anyway, that's fine
17:19
<rbyers>
annevk: Ok, no worries. As long as the fundamental approach (which we're actively implementing now) isn't in question then there's no rush...
17:20
<annevk>
rbyers: I don't think so, though as you can tell I forgot some of the background :/
17:20
<annevk>
(but not the bit where we decided this was the way forward)
17:20
<rits>
annevk: actually i have changed a line from a paragraph, i have tried to make the spaces right without changing other one's, i think it will be fine now
17:21
<rbyers>
annevk: Yeah, that's my fault for dragging the review out <grin>. But it does seem like we keep coming back to the observibility issue though. It's good to hash it out because even once it's in the spec, you won't be the last one to be concerned about it ;-)
17:22
<rbyers>
annevk: I think it's a legitimate viewpoint that passive listeners make the observability problem worse in some ways, but also better in other ways (in particular it gives us an incremental path to potentially solving the problem entirely).
17:23
<rbyers>
... and IMHO the ways in which it makes the problem worse are subtle and unlikely to be an issue for developers in practice, and the ways it makes it better are fundamental and can lead to dramatic improvements in the developer experience :-)
17:23
<annevk>
I guess they mostly serve as a reminder that you don't want to design any new event system in a way that would require them
17:24
<annevk>
And by "new event system" I mean a new set of events to expose some functionality not exposed now
17:24
<rbyers>
... Right. The real problem IMHO is that we've avoided the complexity in the spec by not discussing this in the past (even though it's been the reality since the iPhone release). In order to fix the problem, we have to describe / expose it, which adds complexity to the spec (in order to document reality) :-(
17:25
<rbyers>
annevk: Not sure how to best capture all this subtlety in the spec, but happy to keep iterating on it with you until you're happy :-)
17:27
<annevk>
Thanks, I'll see if I can help when I get back. Pretty sure it'll land in (early) January
18:08
<annevk>
rbyers: so technically dictionary members are present or not present per https://heycam.github.io/webidl/#dfn-dictionary
18:08
<annevk>
rbyers: "has" is not really a thing
18:08
<annevk>
rbyers: it perhaps should be though
18:08
<annevk>
rbyers: there's a ton of stuff that could make this easier...
18:08
<annevk>
rbyers: anyway, for January I guess
18:09
<annevk>
rbyers: doesn't seem worth spending too much time on, I might tweak it a bit while merging
18:09
<caitp>
does the pass-by-value thing work in any implementations?
18:11
<rbyers>
annevk: ok, sounds good - thanks!
18:13
<rbyers>
annevk: I could change "If <var>options</var> is a dictionary and has member <code>{{EventListenerOptions/capture}}</code> with value true" to "If <var>options</var> is a dictionary and <code>{{EventListenerOptions/capture}}</code> is present in <var>options</var> with value true"
18:15
<annevk>
rbyers: r+
18:16
annevk
goes to make some dinner
18:25
<rbyers>
annevk: done
18:30
<MikeSmith>
botie, inform zcorpan when you say you're running wattsi directly, you mean you're manually running "wattsi /opt/workspace/html-build/.temp/source-whatwg-complete /opt/workspace/html-build/.temp/wattsi-output /opt/workspace/html-build/.cache/caniuse.json /opt/workspace/html-build/.cache/w3cbugs.csv" where the stuff is the .temp and .cache dirs is what wattsi expects?
18:30
<botie>
will do
19:41
<MikeSmith>
smaug____: about https://github.com/w3c/push-api/issues/177 if a "URL-safe base64 encoding" were defined, I wonder what spec would define it
19:41
<MikeSmith>
or really, what it would say
19:42
<MikeSmith>
the Push API editors must have some idea in mind of what it actually is supposed to mean
19:42
<smaug____>
MikeSmith: I was reviewing implementation for that code and couldn't know what should happen there, so I filed that bug
19:43
<smaug____>
maybe referring to ... let me find the link
19:43
<MikeSmith>
I see
19:43
<MikeSmith>
I would think they must be assuming that implementors would understand what it means
19:44
<smaug____>
maybe referring to https://tools.ietf.org/html/rfc4648#page-7 or something would be the right thing to do
19:44
MikeSmith
looks
19:44
<MikeSmith>
aha
19:45
<MikeSmith>
yeah, "Base 64 Encoding with URL and Filename Safe Alphabet"
19:45
smaug____
is just being a pain-in-the-ass-reviewer and complaining about stuff ;)
19:45
<MikeSmith>
The "URL and Filename safe" Base 64 Alphabet
19:46
<MikeSmith>
well it should be clear it and should use exact precise terms and not synonymns for them
19:47
<smaug____>
if you look at the wikipedia article, it mentions couple of variants of base64url
19:47
MikeSmith
looks
19:48
<MikeSmith>
「Standard 'base64url' with URL and Filename Safe Alphabet」vs 「Unpadded 'base64url' for "named information" URI's」
19:48
<smaug____>
and = handling
19:48
<smaug____>
" Some libraries ... will encode '=' to '.'."
19:49
<MikeSmith>
oh
19:49
<TabAtkins>
smaug____: That's def not a pain-in-the-ass nitpick. I had no idea what that was, people need to link to things more obscure than ASCII. ^_^
20:15
<MikeSmith>
Domenic: OK finally made the fixes to https://github.com/whatwg/html-build/pull/29
20:16
<Domenic>
MikeSmith: yay! will review today
20:16
<MikeSmith>
only took me 80 days!
20:16
<MikeSmith>
Domenic: super, thanks
20:35
<botie>
zcorpan, at 2015-12-17 16:08 UTC, MikeSmith said: here now if you have time to chat about the build script and at 2015-12-17 18:30 UTC, MikeSmith said: when you say you're running wattsi directly, you mean you're manually running "wattsi /opt/workspace/html-build/.temp/source-whatwg-complete
20:35
<botie>
/opt/workspace/html-build/.temp/wattsi-output /opt/workspace/html-build/.cache/caniuse.json /opt/workspace/html-build/.cache/w3cbugs.csv" where the stuff is the .temp and .cache dirs is what wattsi expects?
20:36
<zcorpan>
MikeSmith: manually typing such yes
20:36
<MikeSmith>
zcorpan: ok will try that
20:39
<zcorpan>
for example i get Error: Multiple secondary definitions (subdfn) for term "dom-context-2d-settransform" Parent of first says: "context . setTransform(a, b, c, d, e, f)", parent of second says: "context . setTransform(matrix)" plus 40-something similar errors
20:47
<zcorpan>
though i ran it on source directly, maybe that's why there were errors?
20:49
<MikeSmith>
zcorpan: no, when I can reproduce it when running it manually on the pre-processed source as well
20:49
<zcorpan>
have to go now
20:51
<MikeSmith>
k
20:55
<MikeSmith>
so the errors are getting written to .temp/wattsi-output.txt but not echoed after that
20:57
<MikeSmith>
maybe we should just be using tee there to begin with but I think the reason I did it that way was so that we just do that same thing we do for the remote-wattsi case, where the errors are written to a file an the server side (and then we download that file and cat it after the build completes)
20:58
MikeSmith
will fix it now
21:05
<MikeSmith>
so it seems that wattsi doesn't actually exit non-zero when it hits those errors, and all we're doing it checking for non-zero exit status, and since we don't find it, the script assumes everything's fine and there's nothing to report
21:06
<MikeSmith>
fix would seem to be to make wattsi actually exit non-zero
21:08
<Domenic>
+1
21:09
<Domenic>
Although I guess maybe unix programs are supposed to only exit nonzero when they fail to produce any output?
21:15
<wanderview>
Domenic: I don't think that's the case...
21:15
<wanderview>
or I have never heard of that rule before
21:22
<stakagi>
I want someone to teach the reason for Dimension attributes being not a floating point but integer.
21:23
<stakagi>
html spec. has said that the widthheight attribute of iframe is non negative integer.
21:24
<stakagi>
Firefox and IE actually parse it as integer, on the contrary, chrome and safari parse it as floating point.
21:24
<stakagi>
On the corresponding property on CSS, these are floating points.
21:24
<stakagi>
https://html.spec.whatwg.org/multipage/embedded-content.html#attr-dim-width
21:37
<TabAtkins>
Presumably because they were invented back in the days when nobody could have dreamed of a 2x display, so supporting floating point was pointless. (sorry)
21:40
<stakagi>
I am verifying the embedded content chapter of svg2 and noticed this.
21:41
<stakagi>
I am putting that note into svg2 wd.
21:42
<caitp>
i I had to guess, i'd say it's probably just because webkit's parseSimpleLengthValue() always results in a double
21:43
<caitp>
and blink probably still inherits that property
21:43
<caitp>
but, just guessing, more css-savvy igalians would know more
21:44
<caitp>
mozilla/msft might just cut corners and avoid going to all that trouble
22:06
<TabAtkins>
Moz/IE aren't cutting corners, they're following the spec. ^^_
22:09
<zcorpan>
MikeSmith: i tried to make wattsi's Fail() write to stderr with Writeln(StdErr, ...) but it seems it didn't make any difference when doing e.g. wattsi ... 2>test.txt
22:10
<zcorpan>
MikeSmith: changing the return value still doesn't print the errors, right?
22:12
<zcorpan>
maybe there's something obvious i don't know about in bash where you can print what wattsi prints here https://github.com/whatwg/html-build/blob/master/build.sh#L182
22:16
<zcorpan>
oh, it's just stdout that's written to that file?
22:19
<MikeSmith>
zcorpan: yeah just stdout
22:19
<MikeSmith>
not stderr
22:19
<MikeSmith>
not sure wattsi ever writes anything to stderr
22:21
<zcorpan>
probably not, i was just experimenting with changing how Fail() writes
22:21
<zcorpan>
so what does https://github.com/whatwg/wattsi/blob/master/src/wattsi.pas#L1556 do?
22:27
<zcorpan>
MikeSmith: is there a problem with not redirecting stdout to a file, and just let it print its stuff?
22:55
<MikeSmith>
zcorpan: the problem is that we can't do that for the remote build
22:55
<MikeSmith>
we need to write it to a file on the server side for the remote case
22:56
<MikeSmith>
and the same code shared for the final error reporting is used after the build is run locally or remotely
22:56
<MikeSmith>
and it's agnostic to where the build ran
22:56
<MikeSmith>
it just cats that file
22:57
<MikeSmith>
and the reason for sharing it is that we have to actually run the build twice
22:57
<MikeSmith>
to get the right line numbers
22:57
<MikeSmith>
if/when there are line numbers to repot
22:58
<MikeSmith>
which is not actually the case always because for some errors wattsi reports no line numbers
22:58
<MikeSmith>
but anyway, that's why we need to delay the error reporting
22:59
<MikeSmith>
we could have it report the errors in real time when it finds them in the local case but then if there are line numbers, the numbers are off by hundreds of lines
22:59
<MikeSmith>
because it;s reporting them against the pre-processed source
23:01
<MikeSmith>
obviously it's an ugly hack the only alernative is that we need to teach wattsi how to calculate the offset so that it reports the correct line numbers even when run just once against the pre-processed source (instead of the original raw source)
23:02
<MikeSmith>
and we should probably do that eventually but it would require substantially more work laboring in Pascal code
23:04
<MikeSmith>
given all that if you have a better idea of how to handle the error reporting that doesn't require caching it like this is doing, then that would be great
23:36
<zcorpan>
MikeSmith: ok. but can it print the content of that file even on success?