[Laszlo-dev] Someone broke lztest

[Henry Minsky]

It's not non-deterministic, it's just that the two lines

lz.DeclaredEventClass = LzDeclaredEventClass;  // publish
lz.Formatter = LzFormatter;  // publish

do occur before the lz object is defined.

I'm going to move those two declarations to after the code that
initializes the lz object.



On Tue, Jul 8, 2008 at 3:32 PM, Henry Minsky <hminsky at laszlosystems.com> wrote:
> Something seems to be broken now when a swf9 app starts up.
> I get a runtime error
>
> TypeError: Error #1010: A term is undefined and has no properties.
>        at LFCApplication$cinit()
>        at global$init()[/private/tmp/lzswf9/lzgen37472/LFCApplication.as:11]
>        at global$init()[/Users/hqm/openlaszlo/lib/jakarta-tomcat-5.0.30/temp/lzswf9/lzgen41232/LzApplication.as:11]
>
>
> That line is right at the start of the class definition for the base
> application, LFCApplication, so I assume it
> is an error happening in the static initializers. It looks like the
> "lz" object is references before it is initialized:
>
>
>    public class LFCApplication extends Sprite {
>
> lz.Formatter = LzFormatter;
>
> lz.DeclaredEventClass = LzDeclaredEventClass;
> (function  () {
> if (lz is Object) {
>
> } else {
> if (!lz) {
> lz = new LzInheritedHash()
> } else {
>
> }}})();
>
> lz.Eventable = LzEventable;
> lz[LzNode.tagname] = LzNode;
> lz.Node = LzNode;
>
>
> So there is maybe something non-deterministic about when the compiler
> emits that initializer function for the "lz" object?
>
> And it does seem redundant; do we really need "lz.node" AND "lz.Node"?
>
>
>
>
>
>
> On Tue, Jul 8, 2008 at 2:44 PM, P T Withington <ptw at pobox.com> wrote:
>> 'point release' is confusing terminology.
>>
>> See my email to laszlo-dev on backward-compatibility policy.  I think it is
>> ok for us to eventually remove the old names on a minor release, but we have
>> to first have a release where the old names cause a deprecation warning.  So
>> the bug here is that the old names have been removed without a deprecation
>> warning, hence you get a fatal error.
>>
>> I know we deprecated the old names in the doc, but I don't think that is
>> sufficient.  It has to be the case that my code gets (preferably) a compiler
>> or runtime warning for using the deprecated name and I can transition to the
>> new way to stop the warning.
>>
>> We either need to reopen the bug associated with r10175 or file a new bug to
>> restore the deprecated names and issue warnings for this release.
>>
>> On 2008-07-08, at 08:29 EDT, Henry Minsky wrote:
>>
>>> yeah, we ought to keep the old names until the next major release.
>>>
>>>
>>>
>>> On Tue, Jul 8, 2008 at 7:38 AM, P T Withington <ptw at pobox.com> wrote:
>>>>
>>>> So, we are removing the old global names for services on a point release?
>>>> That seems wrong to me.
>>>>
>>>> On 2008-07-07, at 23:27 EDT, Henry Minsky wrote:
>>>>
>>>>> it was a bug in the runlztest.sh script that runs the tests in rhino;
>>>>> a reference to LzIdle needed to be
>>>>> changed to lz.Idle
>>>>>
>>>>> fixed in r10245
>>>>>
>>>>> On Mon, Jul 7, 2008 at 7:49 PM, Henry Minsky <hminsky at laszlosystems.com>
>>>>> wrote:
>>>>>>
>>>>>> Changeset  10175 (philr's big renaming changeset) is when it broke.
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Mon, Jul 7, 2008 at 5:17 PM, P T Withington <ptw at pobox.com> wrote:
>>>>>>>
>>>>>>> That would be great if you could.  It just started happening last
>>>>>>> Friday,
>>>>>>> AFAIK.
>>>>>>>
>>>>>>> On 2008-07-07, at 17:04 EDT, Henry Minsky wrote:
>>>>>>>
>>>>>>>> Is anyone looking at this?  If not, I'll try a binary search to see
>>>>>>>> when it broke.
>>>>>>>>
>>>>>>>>
>>>>>>>> On Thu, Jul 3, 2008 at 9:35 AM, P T Withington <ptw at pobox.com> wrote:
>>>>>>>>>
>>>>>>>>> [exec] js: "tmp/test_lztest_lztest-static-attrs.js", line 84:
>>>>>>>>> uncaught
>>>>>>>>> JavaScript runtime exception: TypeError: Cannot call method
>>>>>>>>> "sendEvent"
>>>>>>>>> of
>>>>>>>>> undefined
>>>>>>>>>
>>>>>>>>> On 2008-07-03, at 05:57 EDT, svnbuild at laszlosystems.com wrote:
>>>>>>>>>
>>>>>>>>>> FAILURE: build failed (1) for 10186-openlaszlo-trunk at Thu Jul 3
>>>>>>>>>> 02:57:04
>>>>>>>>>> PDT 2008 (Megatest failed) attaching
>>>>>>>>>> /home/svnbuild/logs/tail-10186-trunk-nightly.log
>>>>>>>>>> <tail-10186-trunk-nightly.log>
>>>>>>>>>> _______________________________________________
>>>>>>>>>> Laszlo-builds mailing list
>>>>>>>>>> Laszlo-builds at openlaszlo.org
>>>>>>>>>> http://www.openlaszlo.org/mailman/listinfo/laszlo-builds
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> --
>>>>>>>> Henry Minsky
>>>>>>>> Software Architect
>>>>>>>> hminsky at laszlosystems.com
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Henry Minsky
>>>>>> Software Architect
>>>>>> hminsky at laszlosystems.com
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Henry Minsky
>>>>> Software Architect
>>>>> hminsky at laszlosystems.com
>>>>
>>>>
>>>
>>>
>>>
>>> --
>>> Henry Minsky
>>> Software Architect
>>> hminsky at laszlosystems.com
>>
>>
>
>
>
> --
> Henry Minsky
> Software Architect
> hminsky at laszlosystems.com
>



-- 
Henry Minsky
Software Architect
hminsky at laszlosystems.com