Writing self-executing functions with !function(){}() is sometimes unsafe

[brycecr]

Consider the example:

(function(){ doSideEffects(); return 5; })() // A

versus

!function(){ doSideEffects(); return 5; }() // B

A and B cause the same side effects, but the result of each expression is different, 5 in A and false in B

I am uglifying code files that look something like A, which are rewritten to B, and I am running into this issue because the minified files are passed to a processor that uses the result of eval() on the minified code.

I did not find a flag to uglify that controls this behavior, but I was curious whether:

• Such a flag exists
• There is a way around this problem without hacking around uglify
• There should be a flag (or patch) for this
• Other thoughts of people who understand Uglify better than I do

[mishoo]

There is a flag, try passing -c negate_iife=false.

[michaelficarra]

That is not the right answer. You are looking to preserve the program completion value. See #543 for the previous discussion about this.

[mpcref]

I propose the following fix:
never negate an IIFR that is also the program completion value.
New issue #628 created.

[rvanvelzen]

Track #640 for follow-up.