Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AwSnap – This link crashes Chrome

github.com

Post reply on HN
Print view View original

21–30 of 43 posts

Re: AwSnap – This link crashes Chrome

#21

Post by jlblatt »

greggman wrote: post #20

I hope this doesn't come across as negative but why post this on HN? Why not just file a bug on crbug.com? Browsers have bugs, all them. It seems kind of negative to turn it into a spectacle rather than just be kind and report the bug to the team.

Totally valid, and not negative at all. See: https://news.ycombinator.com/item?id=9326370

Re: AwSnap – This link crashes Chrome

#24

Post by jlblatt »

myworldplz wrote: post #23

I've created something similar but for Firefox: http://crashfirefox.com/ Have added new crash options for users with NoScript (even if you're blocking 302s - try right clicking).

Hah, that click was definitely my karma for today, ty.

But the reason I posted this is because it doesn't rely on Javascript to eat up memory- anyone can post a malformed/long link to a web forum and crash the thread/site for other Chrome users immediately, without clicking thru.

Re: AwSnap – This link crashes Chrome

#25

Post by nmjohn »

myworldplz wrote: post #23

I've created something similar but for Firefox: http://crashfirefox.com/ Have added new crash options for users with NoScript (even if you're blocking 302s - try right clicking).

And it just keeps going, love it

    ➜  ~  curl http://crashfirefox.com/

    ....................../´¯/)
    ....................,/¯../
    .................../..../
    ............./´¯/'...'/´¯¯`·¸
    ........../'/.../..../......./¨¯\
    ........('(...´...´.... ¯~/'....')
    .........\.................'...../
    ..........''...\.......... _.·´
    ............\..............(
    ..............\.............\...
        SNITCHES GET STITCHES

    ....................../´¯/)
    ....................,/¯../
    .................../..../
    ............./´¯/'...'/´¯¯`·¸
    ........../'/.../..../......./¨¯\
    ........('(...´...´.... ¯~/'....')
    .........\.................'...../
    ..........''...\.......... _.·´
    ............\..............(
    ..............\.............\...
        SNITCHES GET STITCHES

    ....................../´¯/)
    ....................,/¯../
    .................../..../
    ............./´¯/'...'/´¯¯`·¸
    ........../'/.../..../......./¨¯\
    ........('(...´...´.... ¯~/'....')
    .........\.................'...../
    ..........''...\.......... _.·´
    ............\..............(
    ..............\.............\...
        SNITCHES GET STITCHES


    ➜  ~  curl http://crashfirefox.com | wc -l
      % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                     Dload  Upload   Total   Spent    Left  Speed
    100 3613k    0 3613k    0     0  2020k      0 --:--:--  0:00:01 --:--:-- 2019k
      120000

Re: AwSnap – This link crashes Chrome

#26

Post by ploxiln »

greggman wrote: post #20

I hope this doesn't come across as negative but why post this on HN? Why not just file a bug on crbug.com? Browsers have bugs, all them. It seems kind of negative to turn it into a spectacle rather than just be kind and report the bug to the team.

Browser makers need pressure to prioritize robustness. Otherwise, to be competitive with each other, they will prioritize features much more highly, and we can see the results today. "Browsers have bugs, all of them." ...but we're not talking about very-tricky-to-trigger bugs found a couple of times a year, we're talking about 30 every 6 weeks, in various subsystems.

I was initially turned off from Chrome in its early versions because it seemed that tabs crashed much more easily, since it wasn't as big a deal to have a tab crash. Firefox crashes as a whole, so it has much more motivation to diligently make sure its subsystems are not prone to crashing on high-level documents / languages, however malformed they may be.

But as was posted below, it's still quite easy to lock up Firefox. So I applaud crashfirefox.com as well as this Chrome demonstration, they're needed to keep the balance between robustness and new features.

Re: AwSnap – This link crashes Chrome

#27

Post by ploxiln »

nmjohn wrote: post #25
myworldplz wrote: post #23

I've created something similar but for Firefox: http://crashfirefox.com/ Have added new crash options for users with NoScript (even if you're blocking 302s - try right clicking).

And it just keeps going, love it ➜ ~ curl http://crashfirefox.com/ ....................../´¯/) ....................,/¯../ .................../..../ ............./´¯/'...'/´¯¯`·¸ ........../'/.../..../......./¨¯\ ........('(...´...´.... ¯~/'....') .........\.................'...../ ..........''...\.......... _.·´ ............\..............( ..............\.............\... SNITCHES GET STITCHES ......................…

Hah, I did the same thing, but it's not too hard to get past:

    curl -i -H 'User-Agent: Firefox' crashfirefox.com | head -n 80 | cut -b 1-100

Re: AwSnap – This link crashes Chrome

#28

Post by Smerity »

ahoge wrote: post #16
Smerity wrote: post #15

Tangentially related, I was performing an analysis of the billions of links available in Common Crawl[1] and ran across an interesting error: there was one link where the port number was 18779690999[2]. The library I was using converted the port text into a 32 bit integer but the library didn't account for the possibility that someone supplied a port well over the 32 bit int limits. I was curious how many tools that…

The port number is supposed to be a 16-bit number though. So, anything above 65535 is malformed. http://en.wikipedia.org/wiki/Port_%28computer_networking%29

Agreed that it's malformed. My point was more that I'm curious how many other libraries fail badly when confronted with an unexpectedly large port. It's the edge cases / "no-one would ever be silly enough to do that" that quite frequently lead to security issues =]

Re: AwSnap – This link crashes Chrome

#29

Post by jlblatt »

Smerity wrote: post #28
ahoge wrote: post #16

Earlier quoted context omitted.

The port number is supposed to be a 16-bit number though. So, anything above 65535 is malformed. http://en.wikipedia.org/wiki/Port_%28computer_networking%29

Agreed that it's malformed. My point was more that I'm curious how many other libraries fail badly when confronted with an unexpectedly large port. It's the edge cases / "no-one would ever be silly enough to do that" that quite frequently lead to security issues =]

Yes, that's why I feel like this is more significant than a simple crash at first glance- no one would purposefully craft a 256+ unchecked character URL, unless they are being malicious, in which case they absolutely will craft a 256+ unchecked character URL.

Re: AwSnap – This link crashes Chrome

#30

Post by userbinator »

jlblatt wrote: post #29
Smerity wrote: post #28

Earlier quoted context omitted.

Agreed that it's malformed. My point was more that I'm curious how many other libraries fail badly when confronted with an unexpectedly large port. It's the edge cases / "no-one would ever be silly enough to do that" that quite frequently lead to security issues =]

Yes, that's why I feel like this is more significant than a simple crash at first glance- no one would purposefully craft a 256+ unchecked character URL, unless they are being malicious, in which case they absolutely will craft a 256+ unchecked character URL.

URLs can definitely get rather long if they contain lots of query data, and the de-facto limit seems to be around 2KB[1]. It's still not acceptable for such URLs to crash the browser though.

[1] http://stackoverflow.com/questions/417142/what-is-the-maximu...

Post reply on HN