Well, MStringLength::foldsTo won't be called unless you have a constant string and you are explicitly using the length property of this string somewhere in the code. Ex: *var s ="some str"; var l = s.length;*
The bug report is actually asking to simplify concatenation of constant strings. In fact, your test case (*var s = "hoge" + x + "fuga";*) is not a target for this bug, since you don't know nothing about *x* at compile time. If you use iongraph to produce the code for the test case in the bug report (*var s = "f" + "oo" + x + "b" + "ar";*), at some point you will see a instruction like "concat constant constant". Since the operands are both constants, you can simplify this instruction at compile time to a simple "constant string", resulting from the concatenation of the two constant operands. As the report says, to do so, you need to write a new *foldsTo*method for the MConcat class to handle this simplification. The *foldsTo* method of the MStringLength class is a good base example. On Thu, Apr 4, 2013 at 9:47 AM, 井関正也 <[email protected]> wrote: > Now, I understand how to call foldsTo function. > > But I can't still fix this bug( > https://bugzilla.mozilla.org/show_bug.cgi?id=852791) > When I type: > var s = "hoge" + x + "fuga" > ,ValueNumberer::simplify or ValueNumberer::simplifyControlInstruction > always call MDefinition::foldsTo(bool useValueNumbers). > > I predict MStringLength::foldsTo is called when string constants is folded. > But MStringLength::foldsTo isn't call. > > What function call when string constants is folded. > _______________________________________________ > dev-tech-js-engine-internals mailing list > [email protected] > https://lists.mozilla.org/listinfo/dev-tech-js-engine-internals > -- Péricles <http://dcc.ufmg.br/%7Epericlesrafael> _______________________________________________ dev-tech-js-engine-internals mailing list [email protected] https://lists.mozilla.org/listinfo/dev-tech-js-engine-internals
