515380 - TM: parseInt giving inconsistent results with JIT on 1.9.1

Status: RESOLVED DUPLICATE of bug 507993

(Core :: JavaScript Engine, defect)

Description

[hanguofeng]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.2; zh-CN; rv:1.9.1.2) Gecko/20090729 Firefox/3.5.2 (.NET CLR 3.5.30729)
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.2; zh-CN; rv:1.9.1.2) Gecko/20090729 Firefox/3.5.2 (.NET CLR 3.5.30729)

Ok,let’s get to the codes below:

  for(var i=0;i < 1000;i++)
  {
        var result = parseInt(-15.02);
	document.write( result );
	document.write( '' );
  }

What do you think it will output?Rows of -15?It might be in other browsers except Firefox 3.5.2 according to my test.In Firefox 3.5.2,only the begins are -15,others are -16,and the number of -15 is varying at different time,sometime 3 and sometimes maybe 4.

Then,codes following will works:

parseInt(-15.02,10);

Reproducible: Always

Steps to Reproduce:
Ok,let’s get to the codes below:

  for(var i=0;i < 1000;i++)
  {
        var result = parseInt(-15.02);
	document.write( result );
	document.write( '' );
  }

What do you think it will output?Rows of -15?It might be in other browsers except Firefox 3.5.2 according to my test.In Firefox 3.5.2,only the begins are -15,others are -16,and the number of -15 is varying at different time,sometime 3 and sometimes maybe 4.

Then,codes following will works:

parseInt(-15.02,10);
Actual Results:  
-15
-15
-15
-16
-16
....(all -16)

Expected Results:  
-15
-15
...(all -15)

Comment 1

[Dave Garrett]

Reproduced with Firefox 3.5.3 and latest Shiretoko on Linux. However, I can't reproduce this with Namoroka or Minefield. Looks like it was fixed somewhere but the fix hasn't been ported over.

Comment 2

[Dave Garrett]

Attachment: testcase

Comment 3

[Dave Garrett]

This is JIT only, by the way. Disabling javascript.options.jit.content in about:config fixes it.

Dupe of bug 507993?

Comment 4

[Andreas Gal :gal]

Thanks for the triaging Dave. #3 is correct, this is a dup.

shell testcase:

for (var i = 0; i < 10; ++i)
    print(parseInt(-15.02));

This is WFM on TM tip. However, bug 507993 wasn't approved to go into 3.5.x. I think this is a bad enough correctness issue that we should take it, so I will request approval.