Parsing can hang on attributes with many references

[rossj]

I've come across a few XML documents that seems to stop my program in its tracks. I've traced the issue down to an issue with the decoding of many reference characters in a single attribute value.

For example, take the following example XML with 35 references:

<a b="&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;"></a>

As is, it takes over 9 minutes to parse on my machine. Decreasing to 34 <, it takes 4.8 minutes, while increasing to 36 gives a parse time of 17.5 minutes. My guess is that this is due to a "catastrophic backtracking" RegExp.

Interestingly, self-closing the <a> tag makes things work at a normal / expected speed.

<a b="&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;"/>