Browser Bugs

This table does not purport to list even all the bugs that we work around in the library, let alone all the many CSS and DOM bugs that exist. Rather this just lists a few browser bugs that are particularly egregious or hard to work around cleanly, and so affect our ability to support the browser with our library.

For some general discussion of available browsers, see browsers.html .

All of these tests pass on: Mozilla 1.4a, IE6, Opera 7.5. It is only the KHTML browser family that is still shipping product with such basic bugs at the ECMAScript language level.

Oh, and guess what? I can crash the entire Safari 1.2 browser (click to make it happen):

crash_safari()

View source for more information.

These bugs have been reported to Apple as Bug ID# 3656443. They have also been reported to KDE as Bug 81728 .

An "X" means it has the bug.

burst

Bug	Click to Test	IE5.0PC	IE5Mac	Saf 1.x	Konq 3.x	Op 7.10	ICE 5.4.1
no "instanceof" operator	`if ({} instanceof Object) {} alert('ok')`		X
no boolean "in" operator	`if ('a' in window) {} alert('ok')`	X	X
problem parsing implied semi-colon	`try{throw new Error('boo')} catch(e) {} alert('ok')`			X	ok in 3.2, X prior
problem parsing named functions	`function noop(f) {}; noop(function foobar() {}); alert('ok')`			X	X
missing Node.addEventListener and Node.attachEvent	`if (document.body.addEventListener \|\| document.body.attachEvent) alert('ok')`		X
missing Document.createEvent and Document.createEventObject	`if (document.createEventObject) alert('ok'); else if (document.createEvent && (typeof document.createEvent('MouseEvents').initMouseEvent == 'function')) alert('ok')`	X	X	X	?
can't load a script via createElement('script') (all can do a document.write creation before onload)	`var el = document.createElement('script'); el.src = 'bb_script.js'; document.body.appendChild(el);`		X	X	X	ok in 7.5, X prior	X
missing encodeURIComponent and decodeURIComponent	`encodeURIComponent('a'); decodeURIComponent('a'); alert('ok')`	X	X	ok in 1.2, X prior	ok 3.2, X prior
missing Number.toFixed (and toPrecision, toExponential)	`(.51).toFixed(0); alert('ok')`	X	X	X	ok in 3.2, X prior
inconsistent Math.floor and Math.round on large numbers	`var n = 1000000000000000; if (Math.floor(n+0.5) == Math.round(n)) alert('ok')`			X	X
no String.replace(RegExp, Function)	`if (("aaa").replace(/(\w)/g, function() {return 'b'}) == 'bbb') alert('ok');`	X	X	X	ok in 3.2, X prior
setTimeout of a function	`setTimeout(function() {alert('ok')},0)`			ok in 1.2, X prior	ok in 3.2, X prior
buggy Array.concat	`if (1 == [].concat([['a']][0]).length) alert('ok')`				ok in 3.1, X prior
return out of some nested switch statements	`/* view source to see test_switch() */ if (test_switch()) alert('ok')`				ok in 3.1, X prior

Dynamic script Support

The W3 standards say nothing about an "onload" handler for script elements (or img elements for that matter: just document, frames, and object). In practice, using the test listed in the table above, here is what we have found for dynamically created script elements:

Browser	script itself	expando = String	setAttribute String	addEventListener
Mozilla 1.7b	Y	Y	Y	Y
IE5.0 Win	Y
Opera 7.5	Y
Safari 1.2	(doesn't work at all)

Dynamic iframe Support

For iframe support, the situation is complicated because the standards are not always clear, and there are many issues. This test and table doesn't even cover all the things that affect us (such as whether "display: none" disables an iframe).

For iframe-based data channel with an onload event independent of the particular url and its content, a browser needs to have support for at least one of:

12,67 (createElement, with onload on iframe element and later replace). ideal, but only Mozilla and Opera 7.5 support it.
ABC (innerHTML, with onload on iframe element). Works on IE6 and Mozilla and Opera 7.5.
5 or Y (document.write with nested iframe). Works on IE6, IE5, Mozilla, Opera 7.x, Safari 1.0. (Will not work on Safari 73 despite passing "5-", because it can't find its parent.)

The above criteria still leave out Konqueror and ICE family browsers. Note that KDE recently added support for W3 "Load and Save", which means that an iframe data channel is not as critical for Konqueror. (See Peter Kelley, http://marc.theaimsgroup.com/?l=kfm-devel&m=105618985421366&w=2 ). Also, Safari recently added XMLHttpRequest support.

To see the source of the test (what each of the letter codes for passes mean) do view source. A quick partial guide:

   1   iframe element onload (element created with createElement)
   A   iframe element onload (element created with innerHTML)
   3   iframe window onload with no src
   5-  iframe body onload (body created with document.write)
   5+  iframe body onload (overridden programmatically)
   Y   nested iframe element onload, from element in iframe body (created with document.write)
   7   iframe element onload after replace

To run the test:

click here to run test_iframe()

Family	Brand Version	Passes	Comment
IE	6.0	BAC82945-XY69	Notice 9 twice. with slight changes to test, can get BAC82465+XY9
IE	5.0	BC5-X69
Mozilla	Mozilla 1.4a	CBA28143XY5-3697	perfect except for 5+ vs. 5-
Opera	7.10	BC8245-X69	only B8 if no previous alert or setTimeout. security error if write to winempty.
Opera	7.5	CB284XY65+9A1717	note duplicate 1 and 7.
KHTML	Safari 1.0	C245-6B89	parent is bad in 5- for v73 but ok in 1.0. Both ok in test 8. crashes if write to winstatic.
KHTML	Safari 1.2	C245-6XY8197
KHTML	Konqueror 3.1.0	CB28	only finds window.frames by index.
ICE	5.4	BC824X69	only B8 if no previous alert or setTimeout