[Edbrowse-dev] 3.6.0?

[Edbrowse-dev] 3.6.0?

[Chris Brannon]

Hi,
I haven't seen anything on this list for a while.
We've made a lot of big changes, but everything seems to be working
pretty well.  Are there any outstanding issues?  Could we do a new
release in the next week or two?

-- Chris

[Edbrowse-dev] 3.6.0?

[Karl Dahlke]

I think Kevin has a couple small patches which will allow jquery to compile,
which would be great, so I'd say after those patches let's lock in
all our hard work under 3.6.0.
We should add a comment to README on the minimum version of tidy
needed for edbrowse to operate properly.

Karl Dahlke

Re: [Edbrowse-dev] 3.6.0?

[Kevin Carhart]

Coming soon.  Coming this evening.


On Wed, 2 Dec 2015, Karl Dahlke wrote:

> I think Kevin has a couple small patches which will allow jquery to compile,
> which would be great, so I'd say after those patches let's lock in
> all our hard work under 3.6.0.
> We should add a comment to README on the minimum version of tidy
> needed for edbrowse to operate properly.
>
> Karl Dahlke
> _______________________________________________
> Edbrowse-dev mailing list
> Edbrowse-dev@lists.the-brannons.com
> http://lists.the-brannons.com/mailman/listinfo/edbrowse-dev
>

--------
Kevin Carhart * 415 225 5306 * The Ten Ninety Nihilists

Re: [Edbrowse-dev] 3.6.0?

[Chris Brannon]

Kevin Carhart <kevin@carhart.net> writes:

> Coming soon.  Coming this evening.

It's all good, I don't think anyone's in a hurry, and this sounds like
some awesome work!

-- Chris

Re: [Edbrowse-dev] 3.6.0?

[Kevin Carhart]

Thanks Chris!

Here is the teaser.  In the just-submitted-to-Karl, if
I compile it right now, I can load a page with jquery
1.9.1, then go to edbrowse jdb mode and echo the jquery
object and it returns the following, which is the verification
that jquery found all its mandatory prerequisites and 'compiled':


---
function ( selector, context ) {
// The jQuery object is actually just the init constructor 'enhanced'
return new jQuery.fn.init( selector, context, rootjQuery );
}
---

And then if I do the object keys enumerator on that, it returns the whole
long litany of jquery methods and properties that pages may be expecting
to find:

---
ok($)

fn,extend,noConflict,isReady,readyWait,holdReady,ready,isFunction,
isArray,isWindow,isNumeric,type,isPlainObject,isEmptyObject,error,
parseHTML,parseJSON,parseXML,noop,globalEval,camelCase,nodeName,
each,trim,makeArray,inArray,merge,grep,map,guid,proxy,access,now,
Callbacks,Deferred,when,support,cache,expando,noData,hasData,
---

And it goes on like that for several more lines!

There are later versions of jquery than 1.9.1, which introduce yet another 
cute library-within-a-library called Sizzle.  We don't have that 
yet, but hopefully the present work will open up a swath of pages!

Kevin

[Edbrowse-dev] 3.6.0?

[Karl Dahlke]

Ok, this is good.
Kevin I installed your latest patch, tested jsrt etc,
and then pushed it to Master,
but I don't think removeEventListener is quite right.
I'm too tired to look at it now but I'll review it tomorrow.
I think you want to compare the passed in function, just as a function,
with the functions in the array, and if found use splice to remove it,
or if it matches the original function then remove that,
or do nothing if it doesn't match any of these.
I doubt you want to expand both functions back into source and compare those,
but hey I'm not sure.
So on we go.

Karl Dahlke

Re: [Edbrowse-dev] 3.6.0?

[Kevin Carhart]

Great!  This is tentatively exciting.  Thanks for fixing 
removeEventListener, Karl.

It hasn't really hit me yet that edbrowse now has basic jquery and this 
can hopefully go into 3.6.0.

It remains to be seen if there will be a payoff for users right now, or if 
it mostly lays the groundwork for newer versions like 1.11 and 2.1.0.

And how much work is the differential between what we have and the latest? 
I'm not sure, I'll find out...

Kevin

[Edbrowse-dev] 3.6.0?

[Karl Dahlke]

> It remains to be seen if there will be a payoff for users right now, or if
> it mostly lays the groundwork for newer versions like 1.11 and 2.1.0.

I suspect the latter.
Complex javascript has to work perfectly from start to end,
or it just doesn't work.
But this foundation is a big step forward.

Karl Dahlke

[Edbrowse-dev] jquery v2 and an interesting seg fault

[Kevin Carhart]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 2765 bytes --]


I found out some interesting things about compiling later versions of
jquery.
Apparently what separates jquery 1.9 support from support for the latest 
is not very much.  I found that I could even get 2.1.4 to compile
if I add three more initializations to the startwindow file:

document.nodeType = 9;
document.implementation = {};
document.location = {};

Apparently these things need to be there for versions of
jquery from 1.10 and upwards to work.

In this thread beginning at around line 165, the Sizzle developer
lays out what is the minimum DOM implementation to "load Sizzle
without exceptions - invocation being another story though".

https://github.com/jquery/sizzle/issues/343

So I'm comparing what we already have with what this person
says is the minspec, and that's how I arrive at needing
nodeType, implementation and location, at least to start,
and get over the catch-22's.

I added these three lines to startwindow and recompiled.
It appears that a lot of "Can't load jquery" runtimes went away.

If these three lines are all benefit and no drawback, I'd
like to send a patch and get them in.  However, I tested on
fanfiction.net, to take a popular scenario, and I got a
seg fault with the following backtrace:

---

(gdb) bt

#0  0x000000000042b7ff in javaSetsLinkage (after=1 '\001', type=114 'r',
     p_j=0x7f8ea71588e0, rest=0x1456250 "div 0x7f8ea7158910,textarea 0x0, 
")
     at html.c:2352

#1  0x0000000000427366 in runScriptsPending () at html.c:663

#2  0x00000000004276c0 in htmlParse (
     buf=0x14417f0 "\264r\304\037D\005\267\030\022\333\061²Uz\231Æ
\224t\225\t6J\325*?\361K\032#\350\061c\341\260\037\336\364\363\034p%\270\241\320K\303t\031D[\363\002\202W+\016w\031\312,{\213\341\034cK\305M\334*2Øc\334kK\337\070\266\311\353\345\223.W\302\313\326\025f«;\260\267\242Qt\325\344\301Ds\274H\201\275\204e\255\311\060^\266\304\324Z\307\362\251\071Z:\250mF\363$2\024h\215\242\017\221\030\r\274\210\226׳\036\274\001", 
remote=1) at html.c:750

#3  0x00000000004153bf in browseCurrentBuffer () at buffers.c:5401

#4  0x00000000004146b1 in runCommand (
     line=0x698b82 <line.8810+2> "http://fanfiction.net") at buffers.c:5065

#5  0x0000000000414e05 in edbrowseCommand (
     line=0x698b80 <line.8810> "b http://fanfiction.net", script=0 '\000')
     at buffers.c:5237

#6  0x0000000000406946 in main (argc=0, argv=0x7ffcb18e6a30) at 
main.c:1372

(gdb) q

---

I hope the included backtrace is legible.  So my question, to
put it concisely, is can you tell just by reviewing the
backtrace what is going wrong with javaSetsLinkage?  It
appears that javaSetsLinkage is encountering a long series
of escaped characters or binary.  I wonder if I'm causing
this with my 3 new initializations, or reaching an existing
bug.

thanks
Kevin

[Edbrowse-dev] jquery v2 and an interesting seg fault

[Karl Dahlke]

> document.nodeType = 9;
> document.implementation = {};
> document.location = {};

The first two are fine and (theoretically) harmless.
The third is not a good idea, as document.location is set at run time
and is a URL object equal to the url of the page you are browsing.
My latest push adds the first two lines.

The seg fault I cannot reproduce, even with these lines added,
at least not on the fanfiction home page.
Perhaps you traveled down the path a ways and read a book or some such.
Still I get a glimpse of the problem from your backtrace.
Somehow and for some reason js is trying to remove a node
from beneath another node that has no children.
Obviously this should never happen, but it does, or does for you,
so my push allows for that.

Karl Dahlke

Re: [Edbrowse-dev] jquery v2 and an interesting seg fault

[Kevin Carhart]

Bliss.  Thanks!  I was hoping it would ring a bell just from
the bt.  Yes, I don't know how I diverged or got that seg fault,
but it appears we now have the best of both worlds.  I don't
want to jump up and down prematurely, but it's good...


Kevin

[Edbrowse-dev] javascript() in timer side effects

[Kevin Carhart]

I pulled up paypal.com to find out if I could
put the new jquery to use.  Among other kinds
of things going on, I found that one of their
timers gives the following:
timer line 1: ReferenceError: javascript is not defined
timer line 1: ReferenceError: javascript is not defined


I think what it traces back to is the SetTimeout code,
where there is a differentiation based on if (fo).

In the true condition, the side effect I think comes
out like:
t{1000|javascript()|0x7f273eb1fd00|1`~@}

In the false condition, this is like Karl's async.html,
and it succeeds:

t{1000|increment()|0x7f368971dd00|1`~@}

So for one thing, I was wondering if I understand
your intention, Karl.  Are you basically saying
"this is an anonymous function anyhow, and we have
to call it something, so let's call it 'javascript'?"
Or alternately, does 'javascript' have meaning as
  a reserved word here?  Is it related to the syntax
of 'javascript:' with the colon, or no relation?

In case you want a recreation of the problem which
doesn't require hacking up paypal code, I made these:

b http://carhart.net/~kevin/async_karls_original.html
b http://carhart.net/~kevin/async_handler_in_variable.html

The former succeeds and the latter fails.  What can be
done?  I was thinking could we send the variable name itself
in the side effect?  That would be, if the phrasing was like

var blah = function() { ... };

The side effect would then contain t{1000|blah()|

thanks
Kevin

[Edbrowse-dev] javascript() in timer side effects

[Karl Dahlke]

> timer line 1: ReferenceError: javascript is not defined

It looks like I laid a foundation, to compile the function if it is a string,
or simply set the function member if it is already a function,
and put it under TimerObject.onclick, so I could then call TimerObject.onclick
to run the code, no matter how it got set up, but then I never actually
called that onclick function.
Never took the last step.
I was trying to compile every time or something, so now with this small fix
all three cases work for me:

"increment()"   // compile this string
increment   // named function
increment_in_var   // var set to function

Karl Dahlke

Re: [Edbrowse-dev] javascript() in timer side effects

[Kevin Carhart]

Thanks!

Wow, this change reveals just how chatty the major websites are.
At db4, on the login page of paypal, there is a constant stream
of js on a very short interval like '50'.
To experience it in a serialized kind of way is interesting,
different from either the web in a GUI, or from a completely
discrete edbrowse session which is quiet until you trigger it.
Kind of reminds me of monitoring tcpdump or an http proxy, except
all it's doing is run client side code over and over.
(until we open the can of worms of XHR!)





On Sat, 5 Dec 2015, Karl Dahlke wrote:

>> timer line 1: ReferenceError: javascript is not defined
>
> It looks like I laid a foundation, to compile the function if it is a string,
> or simply set the function member if it is already a function,
> and put it under TimerObject.onclick, so I could then call TimerObject.onclick
> to run the code, no matter how it got set up, but then I never actually
> called that onclick function.
> Never took the last step.
> I was trying to compile every time or something, so now with this small fix
> all three cases work for me:
>
> "increment()"   // compile this string
> increment   // named function
> increment_in_var   // var set to function
>
> Karl Dahlke
> _______________________________________________
> Edbrowse-dev mailing list
> Edbrowse-dev@lists.the-brannons.com
> http://lists.the-brannons.com/mailman/listinfo/edbrowse-dev
>

--------
Kevin Carhart * 415 225 5306 * The Ten Ninety Nihilists

[Edbrowse-dev] javascript() in timer side effects

[Karl Dahlke]

Sometimes I fear we're doing lots and lots of work
just to run visual effects that we can't render anyways,
or covertly gather information on our net surfing habits,
or put advertising in front of our face.

Karl Dahlke

Re: [Edbrowse-dev] javascript() in timer side effects

[Kevin Carhart]

I know what you mean about this, except I'm
not sure about "just", as in, solely or exclusively that.
Isn't it, a lot of work in return for useless,
visuals-related objects and methods, but
aggregated together in a bundle
(you have to implement the entire library
or none,) with the
potential for pragmatic and new use cases
which may make the users happy?

I don't know though.  I take your point about
being concerned about where it's heading.

I think the discrete, teletype web is often just
better.  The chatty timers are new, and it's a
little disconcerting to press 'db4' in edbrowse
and realize that "oh, it's no longer entirely
turn based - edbrowse is doing a layer of
things without waiting for me to enable them."
And the more the big websites are working, the more
edbrowse is exposed to what you're talking about,
like advertising and tracking.

I remember reading the list thread earlier this
year that took the example of a stock ticker.
Weren't youall saying,
maybe implement the chatty web in order to
support more sites, but always have an interface
that keeps the user in control with a discrete
'rr' command?  We have 'rr' now- isn't this the
best of both worlds sort of?

> or covertly gather information on our net surfing habits,

Well, one remark on doing things covertly - edbrowse
is very demystifying and a tool for demystification,
I think.
I would say it makes the covert overt.
I think anyone trying to fight tracking has greater power over
internet intrusions to the extent that they flow in in the
form of a discrete CLI or a transactional, matter-of-fact
timestamped log, and the further you get away from the quicksand of
visually oriented design tricks which goes mercilessly
right for your reptilian brain.
For instance, there's the clickbait.
Web advertising uses finely honed images designed
to be as compelling as possible to a viewer's baser
impulses, probably focus-grouped, so that you will
click through.  It works
almost tautologically and is very manipulative.
Then there's the streaming video advertising bloat,
that you are obliged to download and that defaults to "play",
like in the middle of a news story.  I associate this
trick with The Guardian but I'm sure it is a lot of other
places too.  Let's say the Huffington Post, they're
notorious.
Edbrowse is way less covert than this.  I could be
overdramatizing, but I think if
clickbait is a little like "bread and circuses," or the
Roman coliseum, edbrowse restores some
sanity to digitally-mediated environments.

Kevin

Re: [Edbrowse-dev] javascript() in timer side effects

[Adam Thompson]

[-- Attachment #1: Type: text/plain, Size: 4146 bytes --]

On Sat, Dec 05, 2015 at 06:48:10PM -0800, Kevin Carhart wrote:
> 
> I know what you mean about this, except I'm
> not sure about "just", as in, solely or exclusively that.
> Isn't it, a lot of work in return for useless,
> visuals-related objects and methods, but
> aggregated together in a bundle
> (you have to implement the entire library
> or none,) with the
> potential for pragmatic and new use cases
> which may make the users happy?
> 
> I don't know though.  I take your point about
> being concerned about where it's heading.
> 
> I think the discrete, teletype web is often just
> better.  The chatty timers are new, and it's a
> little disconcerting to press 'db4' in edbrowse
> and realize that "oh, it's no longer entirely
> turn based - edbrowse is doing a layer of
> things without waiting for me to enable them."
> And the more the big websites are working, the more
> edbrowse is exposed to what you're talking about,
> like advertising and tracking.

May be but that's where we are. these features are often miss-used,
but discrete web browsing just can't work any more,
at least not if we actually want to remain a useful web browser.

> I remember reading the list thread earlier this
> year that took the example of a stock ticker.
> Weren't youall saying,
> maybe implement the chatty web in order to
> support more sites, but always have an interface
> that keeps the user in control with a discrete
> 'rr' command?  We have 'rr' now- isn't this the
> best of both worlds sort of?

Yeah, I think it's how we have to do things.
Unfortunately, as well as "fun" visuals,
a lot of this stuff is also "functional"
visuals like displaying parts of forms or displaying search results.
One thing I've used (can't remember exactly but I think it was Atlassian Jira)
returned json within an html element then used js to remunge this stuff into
actual search results. There are also other bits of software I regularly
encounter which use timers etc in ways which are tied into the functionality of
the web page and which don't work if ran manually.
For example Rundeck has some sort of timer based scripting to update log files
for executing jobs. Currently this doesn't work properly in Edbrowse (I think
we need AJAX for that as well... that's gonna be fun).

> >or covertly gather information on our net surfing habits,
> 
> Well, one remark on doing things covertly - edbrowse
> is very demystifying and a tool for demystification,
> I think.
> I would say it makes the covert overt.
> I think anyone trying to fight tracking has greater power over
> internet intrusions to the extent that they flow in in the
> form of a discrete CLI or a transactional, matter-of-fact
> timestamped log, and the further you get away from the quicksand of
> visually oriented design tricks which goes mercilessly
> right for your reptilian brain.

Perhaps, but I wonder if we don't need an extension mechanism so that ad
blockers and the like can be created for edbrowse.

> For instance, there's the clickbait.
> Web advertising uses finely honed images designed
> to be as compelling as possible to a viewer's baser
> impulses, probably focus-grouped, so that you will
> click through.  It works
> almost tautologically and is very manipulative.
> Then there's the streaming video advertising bloat,
> that you are obliged to download and that defaults to "play",
> like in the middle of a news story.  I associate this
> trick with The Guardian but I'm sure it is a lot of other
> places too.  Let's say the Huffington Post, they're
> notorious.
> Edbrowse is way less covert than this.  I could be
> overdramatizing, but I think if
> clickbait is a little like "bread and circuses," or the
> Roman coliseum, edbrowse restores some
> sanity to digitally-mediated environments.

I think this is a little over-dramatic tbh but I take your point.
However, we can't not develop features and modernise Edbrowse in an attempt to
ad block because then users just have to use other web browsers and interfaces
or not use most of the internet.

Cheers,
Adam.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 473 bytes --]