Bug 21801 – REGRESSION (r37821): YUI date formatting JavaScript puts the letter 'd' in place of the day

• Component: JavaScriptCore
• Status: RESOLVED FIXED

[Nate Silva]

SUMMARY

When using the YUI DataTable object, a column formatted as:

  {key:'date', formatter:YAHOO.widget.DataTable.formatDate, 
                  sortable:true, resizeable:true}

...shows the date incorrectly. For example, a date which should be shown as "03/24/1980" is instead shown as "03/d/1980".

This does not happen on Safari 3, but it does happen on the latest WebKit nightly (r37764).

STEPS TO REPRODUCE

Go to the Yahoo UI example at:
http://developer.yahoo.com/yui/examples/datatable/dt_basic_clean.html

[Nate Silva]

This has also been filed as YUI bug #2187202.
https://sourceforge.net/tracker2/?func=detail&aid=2187202&group_id=165715&atid=836476

[Alexey Proskuryakov]

Confirmed as a regression with r37807.

[Nate Silva]

Filed with Apple's Bug Reporter (is that what "NeedsRadar" means?). The bug report is:

rdar://problem/6341231

[Alexey Proskuryakov]

Thank you! NeedsRadar isn't necessarily a request for the reporter to file it with Apple (someone else can do that), but it is slightly better if both bugs are technically reported by the same person.

[Cameron Zwarich (cpst)]

I have determined that this regressed between r36766 and r36847. I am also assigning this to myself.

[Cameron Zwarich (cpst)]

There was a typo in my previous post. This regressed in r36821:

http://trac.webkit.org/changeset/36821

[Cameron Zwarich (cpst)]

Created attachment 24962 [details]
Reduction

Here is a reduction. I can't seem to reduce it any further. The code in the script tag also works in the commandline JSC shell.

[Cameron Zwarich (cpst)]

The problem here is that the check in CodeGenerator::emitTypeOf() to determine whether a register is constant is wrong after r36821:

            && static_cast<unsigned>(src2->index()) < m_codeBlock->constantRegisters.size()

I am kind of surprised that nothing else hit this.

[Cameron Zwarich (cpst)]

Created attachment 24967 [details]
Fix

Here is the simple fix. I will make a bunch of tests based on this original pattern, covering all of the individual typeof cases.

[Geoffrey Garen]

In r36821, I had to fix many call sites making custom calculations about whether a register was constant. I guess I missed one.

I think it would be nice to fix this bug by changing this call site to use the isConstant function I wrote -- that way, the next time constant register layout changes, we won't have the same problem again.

[Cameron Zwarich (cpst)]

Created attachment 24972 [details]
Proposed patch

Here is a better fix that uses the isConstant() member function I just added to CodeBlock.

[Geoffrey Garen]

Comment on attachment 24972 [details]
Proposed patch

r=me

[Cameron Zwarich (cpst)]

Landed in r38230.