As Brendan suggested, the "import" and "export" functionality in spidermonkey is legacy stuff from ancient netscape days and is, among other things, consuming a bit of property-attribute space we'd like to free up for properly modeling "const". This patch removes, as delicately as possible, the legacy functionality. I'd appreciate careful review from anyone who feels qualified.

Comment on attachment 331022 [details] [diff] [review] patch to delete the legacy code >+++ b/js/src/jsemit.cpp Wed Jul 23 16:52:50 2008 -0700 >@@ -2393,11 +2393,7 @@ > CG_OFFSET(cg) - pn2->pn_offset) < 0) { > return JS_FALSE; > } >- if (!pn->pn_atom) { >- JS_ASSERT(op == JSOP_IMPORTALL); >- if (js_Emit1(cx, cg, op) < 0) >- return JS_FALSE; >- } else { >+ if (pn->pn_atom) { > if (!EmitAtomOp(cx, pn, op, cg)) > return JS_FALSE; > } No degree of freedom allowing null pn->pn_atom remains, so this code should simplify to one line: return EmitAtomOp(cx, pn, op, cg); (no need to assert pn->pn_atom is not null, we prefer clean null deref crash in general). /be

Comment on attachment 332621 [details] [diff] [review] freshened, reformatted patch with suggested fix Thanks, r=me. /be

Pushed to m-c in revision 216fff17c130.

I'm going to pass creating a test for this. push back if you disagree.

I've been running with this exclusion file for trunk but failed to check it in. Here it is including new exclusions for import|export tests: e4x/Regress/regress-361451.js ecma_2/Exceptions/lexical-010.js ecma_2/Exceptions/lexical-022.js ecma/LexicalConventions/7.4.3-3-n.js js1_5/decompilation/regress-349484.js js1_5/extensions/regress-355622.js js1_5/extensions/regress-418730.js js1_5/GetSet/regress-353264.js js1_5/Regress/regress-249211.js js1_5/Regress/regress-350692.js js1_5/Regress/regress-354924.js js1_5/Regress/regress-362583.js js1_7/extensions/regress-353214-01.js js1_7/lexical/regress-346642-03.js /cvsroot/mozilla/js/tests/spidermonkey-n-1.9.1.tests,v <-- spidermonkey-n-1.9.1.tests initial revision: 1.1 mozilla-central: changeset: 16534:ea3a755dbab5