Bad bug in recent nightly builds

[bob]

The last few recent nightly builds are somehow breaking our SHA256
hashing routines. Here is a test page:

https://rodan.lastpass.com/dev/nightly_problem.php

We aren't sure if some recent javascript engine optimizations are to
blame or what, but if you go to this page in any other browser
(including older builds of chrome), the strings match.

This is a critical bug for LastPass -- it essentially makes our
extension unusable. It obviously has a lot of other potential
implications for any javascript run.

Thanks in advance.

[Drew]

Adding to what Bob mentioned, I've derived a very trivial example from
deep within our SHA256 hashing routines that shows that the latest
Chromium nightlies have some very serious JavaScript problems:

-----

<pre>
<script>
function a()
{
var b;

b = 2600822924;
if (false) {
b = 2600822924;
}
document.writeln('incorrect: ' + (46512102 & b));

b = 2600822924;
document.writeln('correct: ' + (46512102 & b));
}
a();
</script>
</pre>

-----

As you can see, the only difference between incorrect and correct is a
false if statement.

Drew

[Aaron Boodman]

https://rodan.lastpass.com/dev/nightly_problem.php seems to render the
correct thing for me as of r56042 on Linux. What version are you
testing (you can see in about:version).

I'm not sure what I'm looking for with the script snippet Drew sent.
It output this:

incorrect: 33890436 correct: 33890436 which is what it seems like it
should output. Is that not what you're seeing?

- a

> --
> You received this message because you are subscribed to the Google Groups "Chromium-extensions" group.
> To post to this group, send email to chromium-...@chromium.org.
> To unsubscribe from this group, send email to chromium-extens...@chromium.org.
> For more options, visit this group at http://groups.google.com/a/chromium.org/group/chromium-extensions/?hl=en.
>
>

[Robert Billingslea]

I am running 56027 on windows.

https://rodan.lastpass.com/dev/nightly_problem.php consistently produces different values.

When I run Drew's code, I see:

incorrect: 0
correct: 33890436

Bob

[Drew]

I can confirm that I can't reproduce on 56064 under linux64, but I can
still reproduce on 56062 under win32. Haven't tried mac.

Drew

On Aug 13, 1:45 pm, Robert Billingslea <robert.billings...@gmail.com>
wrote:

> I am running 56027 on windows.
>

> https://rodan.lastpass.com/dev/nightly_problem.phpconsistently produces

> different values.
>
> When I run Drew's code, I see:
>
> incorrect: 0
> correct: 33890436
>
> Bob
>
> On Fri, Aug 13, 2010 at 1:31 PM, Aaron Boodman <a...@google.com> wrote:

> >https://rodan.lastpass.com/dev/nightly_problem.phpseems to render the

> > > To post to this group, send email to chromium-extensi...@chromium.org.

> > > To unsubscribe from this group, send email to

> > chromium-extensions+unsubscr...@chromium.org<chromium-extensions%2Bunsubscr...@chromium.org>
> > .

> > > For more options, visit this group at
> >http://groups.google.com/a/chromium.org/group/chromium-extensions/?hl=en.
>
> > --
> > You received this message because you are subscribed to the Google Groups
> > "Chromium-extensions" group.

> > To post to this group, send email to chromium-extensi...@chromium.org.

> > To unsubscribe from this group, send email to

> > chromium-extensions+unsubscr...@chromium.org<chromium-extensions%2Bunsubscr...@chromium.org>
> > .

[Drew]

In further testing, I've found that 56069 under linux32 has the
problem as well.

Drew

On Aug 13, 4:19 pm, Drew <zit...@gmail.com> wrote:
> I can confirm that I can't reproduce on 56064 under linux64, but I can
> still reproduce on 56062 under win32.  Haven't tried mac.
>
> Drew
>
> On Aug 13, 1:45 pm, Robert Billingslea <robert.billings...@gmail.com>
> wrote:
>
> > I am running 56027 on windows.
>
> >https://rodan.lastpass.com/dev/nightly_problem.phpconsistentlyproduces
> > different values.
>
> > When I run Drew's code, I see:
>
> > incorrect: 0
> > correct: 33890436
>
> > Bob
>
> > On Fri, Aug 13, 2010 at 1:31 PM, Aaron Boodman <a...@google.com> wrote:

> > >https://rodan.lastpass.com/dev/nightly_problem.phpseemsto render the

[Arne Roomann-Kurrik]

56036 on Mac exhibits this the undesired behavior as well.

~Arne

To post to this group, send email to chromium-...@chromium.org.
To unsubscribe from this group, send email to chromium-extens...@chromium.org.

[Antony Sargent]

Through the awesome magic of our build-bisect.py script (http://src.chromium.org/viewvc/chrome/trunk/src/build/build-bisect.py?view=log), I have found the culprit to be:

http://src.chromium.org/viewvc/chrome?view=rev&revision=55850

I'm filing a bug and will reply with the link shortly so you can star it to follow progress. 

[Antony Sargent]

http://crbug.com/52096

[Robert Billingslea]

Posted a comment to the crbug report, but want o make sure this gets adequate visibility...

It looks like this bug made it to the dev channel release and we are being hit hard by support emails. Users are frustrated and so are we because we know people are uninstalling because of it.

This morning the bug report was updated to fixed, did that make it into the dev build? If so, then it didn't fix it.

We want to make sure this doesn't make it into the beta release. If you look at the test code in the example provided, it is clear that there are very serious problems on your end.

Thanks again.