Comment #21 on issue 615 by [email protected]: "instanceof" operator broken on x64
http://code.google.com/p/v8/issues/detail?id=615

I know this is going to sound crazy/insane, but I believe I've found another (even weirder) symptom of either this bug, or a similar one, where some kind of registers
mixup is causing the strangest of bugs.

I'm attaching a little code snippet where I have a function that's designed
to "clone" a javascript object recursively. This code works in all browsers, and has also now been tested on Narwhal/Rhino and Node.js on another server. However, of course, it breaks in my regular non-debug V8 build. And yet, predictably, it works
fine in the shell_g debug build on my system

Also, you'll see the code calls the cloneObj function twice, because the bug doesn't
happen the first time, only after the second call.

It should print 3 "0"s since the "links" array should stay empty. Instead, on my busted build, it prints 2 "0"s, and then the most peculiar thing happens. The third print, it prints "first_name". It got so mixed up, that it took the KEY name from the other property and assigned it as the VALUE for the "links" property. Strangest
thing I've ever seen.

I've had my code reviewed several times, and it seems to check out as solid. I can only explain this weirdness by saying it must be related to this registers assembly
mixup that's occuring on my system/build.

-------------

On a side note, since we've proven that at least in some systems (like mine), the debug build and the regular build do NOT function the same, wouldn't it make more sense for V8 to run the test suite with the release build instead of the special debug build? At least in my case, the test suite is completely untrustable because
I've proven now twice that it's not the same as release.

Sure, something about my server must be a rare case (which I hope we someday figure out why), but it would have been nicer if I could at least have run the test-suite and seen some tests fail to KNOW definitively where the failures seem to be, instead of right now, just feeling like I have no idea if I can trust anything about my
build.

Attachments:
        anotherbug1.txt  738 bytes

--
You received this message because you are listed in the owner
or CC fields of this issue, or because you starred this issue.
You may adjust your issue notification preferences at:
http://code.google.com/hosting/settings

--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev

Reply via email to