149801 - Final Release of SpiderMonkey 1.5

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Phil Schwartau]

We want to have zero test failures for the final release of JS1.5.
This meta-bug will be used to track this issue.

Comment 1

[Phil Schwartau]

I currently get these testcase failures on WinNT (debug shell).
One fewer in the opt shell, and two fewer on Linux debug or opt.

REGEXP FAILURES                      (bug 85721)
ecma_2/String/match-002.js
ecma_3/RegExp/regress-123437.js
ecma_3/RegExp/regress-31316.js
ecma_3/RegExp/regress-57572.js
ecma_3/RegExp/regress-85721.js
ecma_3/RegExp/regress-87231.js
js1_2/regexp/vertical_bar.js


OTHER FAILURES
ecma_3/Function/15.3.4.3-1.js        (bug 145791)
js1_5/Array/regress-101964.js        (bug 101964)
js1_5/Object/regress-137000.js       (bug 137000)
js1_5/Object/regress-90596-001.js    (bug 90596)
js1_5/Object/regress-90596-002.js    (  ibid.  )
js1_5/Object/regress-90596-003.js    (  ibid.  )
js1_5/Regress/regress-96128-n.js     (bug 96128)
js1_5/Regress/regress-96526.js       (bug 96526)
js1_5/Exceptions/errstack-001.js


The RegExp failures will all go way once the new RegExp
code is checked in; see bug 85721 for that patch.


The last failure above may be a flaw in my testcase;
let me ask Brendan about that. It fails as follows:

------------------------------------------------------------
js1_5/Exceptions/errstack-001.js failed Bug Number (none)
STATUS: Testing Error.stack
Failure messages were:

FAILED!: Section 7 of test -
FAILED!: Expected value 'A("44:foo","13:bar")@'
         Actual value   'A(44:foo,13:bar)@./js'
------------------------------------------------------------


In Error.stack, it used to be that string arguments to functions
were presented with double-quotes, and this test used to pass.
But now, string arguments are no longer being presented that way.
Is this by design, or is it a bug?

Note that Error.stack was created in response to bug 123177,
and was modified in response to bug 127136.

Comment 2

[Phil Schwartau]

Oops, the SpiderMonkey bug for ecma_3/Function/15.3.4.3-1.js
is bug 145779. The one I listed above is the Rhino analogue.

Note bug 101964 is currently marked as WONTFIX:
"Performance: truncating arrays is slow in SpiderMonkey"

Brendan and I discussed this one; if it isn't too hard
to fix this, it should be fixed for JS1.5

Comment 3

[Phil Schwartau]

Proposing that we get these fixes by [e-mail] in for JS1.5:

bug 14044   jsdtoa.c doesn't check for out-of-memory
bug 148171  js_FinishDtoa() not called unless JS_THREADSAFE is set

Both have patches that need r= and sr=

Comment 4

[Brendan Eich [:brendan]]

Phil, can you add all the bugs listed in this bugs as dependencies ("depends
on") entries in the textfield just under the attachment table?  Then we can
watch as links get crossed out (meaning the bugs are closed).

/be

Comment 5

[Phil Schwartau]

Have added all the above bugs as dependencies, plus bugs in the
JS Engine component that currently have the keyword "js1.5".

Note, we might wish to revisit the js1.5 designation on some of these.

Comment 6

[Kenton Hanson (gone)]

The following is the order I am addressing these bugs,

148171 review of patch done
14044 review of patch nearly done
133548 & 145779 working on patches
96526 future patch
85721 review of patch on hold
96128, 90596 & 61898 future patches
132949 not part of default build
127246 time zone on win2k does not change in js after boot
59659 "might be used uninitialized" compiler warning

Comment 7

[Brendan Eich [:brendan]]

I think that bug 61898 and bug 66381, both of which are optimizations, should
not hold up and put at risk a JS1.5 SpiderMonkey release.  Phil, if you like,
link them to a js1.5optimized release that we can hack on after 1.5 is done.

/be

Comment 8

[Phil Schwartau]

OK, I will file a tracking bug for a js1.5optimized release,
and move those two issues from this tracking bug to the new one.

Comment 9

[Phil Schwartau]

I have filed meta-bug 156155, "JS1.5 Optimized release".

As suggested in Comment #7, I am removing dependencies on bug 61898
and bug 66381 from this bug, and will add them there instead -

Comment 10

[Phil Schwartau]

NOTE: I have just added two new RegExp testcases to the testsuite
that are much larger than usual: 

          mozilla/js/tests/ecma_3/RegExp/perlstress-001.js
          mozilla/js/tests/ecma_3/RegExp/perlstress-002.js

Each one of these contains hundreds of sections, in which we test the
JS RegExp engine against results generated from the Perl 5 engine.

Many sections will fail until the fix in bug 85721 is checked in.
Until then, you will notice a larger-than-normal failure output -

Comment 11

[Kenton Hanson (gone)]

The following is the order I am addressing these bugs,

133548 working on patch
96526 future patch
85721 review of patch on hold
96128, 90596 & 61898 future patches
132949 not part of default build
127246 time zone on win2k does not change in js after boot
59659 "might be used uninitialized" compiler warning

Comment 12

[Kenton Hanson (gone)]

Attachment: bug list

Attached is an Excel spreadsheet exhibiting my prioritization of current js
bugs assigned to me.  I welcome any suggestions or comments.

Comment 13

[Kenton Hanson (gone)]

The current bugs I am working on
140852 & 160602 (floating point) Highest priority.
96128 patch pending
133548
90596
132949

Comment 14

[Yuh-Ruey Chen]

Er, maybe this bug should be renamed :)

Comment 15

[Bob Clary [:bc] (inactive)]

js15 is so out the door.