On Windows, isspace asserts if you give it a negative number (other then EOF?).

1) NS_IS_SPACE(*number) 2) isspace((unsigned char)(*number)) 3) Custom IsSpace like at http://mxr.mozilla.org/mozilla-central/source/content/smil/nsSMILParserUtils.cpp#67 I implemented (1) because it was the first thing I thought of. (3) is probably best because it lets us match the spec. Do SMIL and SVG have the same concept of what is a space character?

White space in lists is defined as one or more of the following consecutive characters: "space" (Unicode code 32), "tab" (9), "line feed" (10), "carriage return" (13) and "form-feed" (12).

http://mxr.mozilla.org/mozilla-central/source/content/svg/content/src/nsSVGNumber.cpp#140 also uses isspace. What definition of "space" does it want? ;)

nsSVGNumber and nsSVGLength want the same definition, that of comment 2. I think there's a function in nsSVGUtils.h for that - IsSVGWhitespace.

Excellent, thanks Robert. IsSVGWhitespace is apparently unused at the moment, but that's easily remedied ;)

Looks good to me. :) FWIW, regarding IsSVGWhitespace being unused -- I initially used IsSVGWhitespace() in the parsing code for <animateMotion>, but I later outsourced that parsing to nsCharSeparatedTokenizer in Bug 562310. And, unfortunately, nsCharSeparatedTokenizer has its own "isWhitespace" method that doesn't catch linefeeds.[1] We probably need to also make nsCharSeparatedTokenizer (and its sibling nsCCharSeparatedTokenizer) have an "SVG mode" to make its isWhitespace method check for linefeeds. That can be a separate bug, though. [1] http://hg.mozilla.org/mozilla-central/annotate/e0c8f11c3892/xpcom/ds/nsCharSeparatedTokenizer.h#l157

Filed Bug 573489 on the nsCharSeparatedTokenizer issue described in comment 7.