[Edbrowse-dev] SegFaults at www.pandora.com

[Edbrowse-dev] SegFaults at www.pandora.com

[Chuck Hallenbeck]

Hi,

I  can access http://www.pandora.com/login.vm without crashing, but
selecting the "register" link on that page, or the top-level address
http://www.pandora.com  both cause an edbrowse segfault. Is this a problem
with a fix? The site uses javascript heavilybut turning js off in edbrowse
only seems to make potential js interactions transparent.

Any ideas appreciated.

Chuck


-- 

Chuck in Hudson. (Hudson on the Hudson)

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Adam Thompson]

Hi Chuck,

On Wed, May 22, 2013 at 06:42:33AM -0400, Chuck Hallenbeck wrote:
> I  can access http://www.pandora.com/login.vm without crashing, but
> selecting the "register" link on that page, or the top-level address
> http://www.pandora.com  both cause an edbrowse segfault. Is this a problem
> with a fix? The site uses javascript heavilybut turning js off in edbrowse
> only seems to make potential js interactions transparent.

Did you ever get any (off-list) responses to this?
I've been trying to replicate the bug,
unfortunately as a UK resident I'm unable to get the same web page.
If the segfault is present with javascript disabled then my next best guess
would be the html parser. What output do you get with debug 7?
Also, what version of edbrowse are you running?

Cheers,
Adam.

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Chuck Hallenbeck]

Hi Adam,

No, I have had no off-list responses to the pandora.com issue. The version
of my edbrowse is 3.4.5, and the crashes are readily reproducible, if you
can get to the site. It seems to be a popular internet radio site for users
of the other OS.

I have uploaded a log of three executions of edbrowse to:

www.hallenbeck.ftml.net/typescript

The first shows the edbrowse version. The second shows the crash with db1,
and suggests a "double free" is probably the cause.

The third execution repeats the second with db7, and is very long. If you
can spot any more information in it that might help, I'd appreciate knowing
about it.

Thanks,

Chuck


-- 

Chuck in Hudson. (Hudson on the Hudson)

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Adam Thompson]

Hi Chuck,

On Fri, May 31, 2013 at 09:23:39AM -0400, Chuck Hallenbeck wrote:
> I have uploaded a log of three executions of edbrowse to:
> 
> www.hallenbeck.ftml.net/typescript
> 
> The first shows the edbrowse version. The second shows the crash with db1,
> and suggests a "double free" is probably the cause.
> 
> The third execution repeats the second with db7, and is very long. If you
> can spot any more information in it that might help, I'd appreciate knowing
> about it.

Thanks. Looking through the last one (with db7)
it looks like the javascript is causing the segfault.
Could you upload a log of the crash with db7 but without javascript?

Cheers,
Adam.
PS: just checked the site again but apparently it's not licensed to broadcast in the UK.

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Chuck Hallenbeck]

Adam,

Without js it does not crash, but the test with db7 is uploaded to:

http://www.hallenbeck.ftml.net/db7nojs.log


Chuck

PS: Interesting about the license restriction. At least that's one mystery
solved.

-- 

Chuck in Hudson. (Hudson on the Hudson)

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Chuck Hallenbeck]

Woops! I forgot to update the web site. The file is there now:

http://www.hallenbeck.ftml.net

Sorry about that.


-- 

Chuck in Hudson. (Hudson on the Hudson)

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Chuck Hallenbeck]

The direct link is:

http://www.hallenbeck.ftml.net/db7nojs.log



-- 

Chuck in Hudson. (Hudson on the Hudson)

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Chuck Hallenbeck]

I'm having a bad day here, sorry. There's a hyphen in the file nme:

http://www.hallenbeck.ftml.net/db7no-js.log



-- 

Chuck in Hudson. (Hudson on the Hudson)

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Chris Brannon]

Chuck Hallenbeck <chuckh@ftml.net> writes:

> Hi Adam,
>
> No, I have had no off-list responses to the pandora.com issue. The version
> of my edbrowse is 3.4.5, and the crashes are readily reproducible, if you
> can get to the site. It seems to be a popular internet radio site for users
> of the other OS.

I reproduced this the other day.  I started trying to debug it, but I
got side-tracked with other things.  Yes, the bug is definitely
JS-related.  When I visit the site, I see a message saying that edbrowse
segfaulted because of a double free or memory corruption.
I'm still looking for the cause...

-- Chris

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Adam Thompson]

On Fri, May 31, 2013 at 10:12:03AM -0400, Chuck Hallenbeck wrote:
> Adam,
> 
> Without js it does not crash, but the test with db7 is uploaded to:
> 
> http://www.hallenbeck.ftml.net/db7nojs.log

Ok, that's what I expected. From glancing through the javascript (from the db7
log with the segfault) I'm wondering if the site's using dynamic options and
it's related to the corresponding bugs mentioned in the documentation (i.e.
the fact that edbrowse doesn't do dynamicly changing pages very well).
I've seen this on more than a few sites (I can't remember them all off the top
of my head), but it seems to be becoming more common as the web evolves.
I'm slowly looking into this (plus a couple of other strange js behaviors such
as the fact that edbrowse occasionally seems to go into an infinite loop when other
browsers are fine).

I'd certainly be interested to know if there's anything I can do as part of a more coordinated effort to help fix this.

Cheers,
Adam.

Re: [Edbrowse-dev] SegFaults at www.pandora.com

[Chuck Hallenbeck]

Hi Adam,

I appreciate your interest in this, and as you may have seen, Chris Brannon
is also looking into it. Hopefully a solution to accessing this particular
site will also be a solution to a larger class of similar ones.

Anything I can do to help, perhaps testing or confirming some observation,
I will gladly do. My technical skills are not up to serious debugging such
things however.

Chuck

On Sat, 1 Jun 2013, Adam Thompson wrote:

> On Fri, May 31, 2013 at 10:12:03AM -0400, Chuck Hallenbeck wrote:
> > Adam,
> >
> > Without js it does not crash, but the test with db7 is uploaded to:
> >
> > http://www.hallenbeck.ftml.net/db7nojs.log
>
> Ok, that's what I expected. From glancing through the javascript (from the db7
> log with the segfault) I'm wondering if the site's using dynamic options and
> it's related to the corresponding bugs mentioned in the documentation (i.e.
> the fact that edbrowse doesn't do dynamicly changing pages very well).
> I've seen this on more than a few sites (I can't remember them all off the top
> of my head), but it seems to be becoming more common as the web evolves.
> I'm slowly looking into this (plus a couple of other strange js behaviors such
> as the fact that edbrowse occasionally seems to go into an infinite loop when other
> browsers are fine).
>
> I'd certainly be interested to know if there's anything I can do as part of a more coordinated effort to help fix this.
>
> Cheers,
> Adam.
>

-- 

Chuck in Hudson. (Hudson on the Hudson)