Apart from your change, I propose some improvements for LzNode:
(after comparing it to the 4.0.8 resp. 3.3.0 versions:
"http://svn.openlaszlo.org/openlaszlo/tags/4.0.8/WEB-INF/lps/lfc/core/LzNode.lzs";
"http://svn.openlaszlo.org/openlaszlo/tags/3.3.0/WEB-INF/lps/lfc/core/LzNode.as";
)
i)
That line is no longer necessary, because the "_instanceAttrs" property
is set earlier. Can be simply removed (after running the standard tests,
for sure :-) ).
// TODO: [2006-05-22 ptw] What is the purpose of this copy?
// construct or constructWithArgs destructively modify args?
var maskedargs = new LzInheritedHash(iargs);
ii)
"_instanceAttrs" should never become "undefined", so we're compliant to
previous versions.
this._instanceAttrs = attrs;
change to:
this._instanceAttrs = attrs || {};
iii)
The same pattern used in "LzNode#_dbg_name()", is also used in
"LzNode#construct(..)"
if (ip.determinePlacement == LzNode.prototype.determinePlacement) {
If that doesn't work for swf9 in "LzNode#_dbg_name()", it won't work
here either.
---
iv)
Just spotted a wrong comment in "LzNode.mergeAttributes(..)":
// Just a normal value (or LzInitExpr), install it
dattrs[k] = attrk;
It will never be a LzInitExpr, because LzInitExpr is handled in the
previous if-condition.
On 4/12/2008 12:32 AM, André Bargull wrote:
Right now I'm doing some additional tests, but I think that fix should
be ok. So, approved from me!
On 4/12/2008 12:22 AM, P T Withington wrote:
I'm going to check this in ahead of André's review, to head off any
additional bug reports (and it's pretty late for him). This fixes a
major error in the commpiling of LZX classes and instances. I am
pretty confident of this fix, because it passes lztest and lzunit.
André: I still look forward to your usual thorough review! :)
On 2008-04-11, at 18:09 EDT, Philip Romanik wrote:
Approved!
Change 20080411-ptw-r by
<mailto:[EMAIL PROTECTED]>[EMAIL PROTECTED] on
2008-04-11
17:21:45 EDT
in /Users/ptw/OpenLaszlo/ringding-clean
for
<http://svn.openlaszlo.org/openlaszlo/trunk>http://svn.openlaszlo.org/openlaszlo/trunk
Summary: Restore instance attributes passed to construct methods
Bugs Fixed:
LPP-5732: 'Regression in Amazon @ 8592'
LPP-5762: 'InstanceAttrs without setters'
Technical Reviewer:
<mailto:[EMAIL PROTECTED]>[EMAIL PROTECTED] (pending) QA
Reviewer:
<mailto:[EMAIL PROTECTED]>[EMAIL PROTECTED]
(pending)
Details:
LzNode: I think I finally understand the dance between "class
instance attributes" and "instance instance attributes". The
attributes put in a class by the tag compiler should be installed
by the constructor, unless they have a setter (which the compiler
could not know about). But attributes declared in an instance are
expected to be passed in 'args' to all construct methods, because
those methods may inspect/handle/modify them before they are
applied. Ugly, fragile mechanism, but we must preserve it for
back-compatibility for the time being. Also revert the change in
placement computation to look for `ignoreplacement` in attrs once
again. Fix up toString and _dbg_name to work without needing
access to .prototype (and to print the `tagname`, not the
`classname`, since the user will be more familiar with the
tagname.
runlzunit: Run the browser in the background on OS X, so I can do
other work while I test.
lztestmanager: Only turn of logging for swf9, otherwise `ant
runlzunit` will always fail!
build.xml: Default swf tests to swf8.
Tests:
ant lztest, and runlzunit, amazon, test case from LPP-5762
Files:
M WEB-INF/lps/lfc/core/LzNode.lzs
M build-tools/runlzunit.sh
M lps/components/lztest/lztestmanager.lzx
M lps/components/base/basetree.lzx
M build.xml
Changeset:
<http://svn.openlaszlo.org/openlaszlo/patches/20080411-ptw-r.tar>http://svn.openlaszlo.org/openlaszlo/patches/20080411-ptw-r.tar
