Bug #35894 for XML-Diff: "Invalid expression" error from XML::Diff

We've observed cases where running an XML::Diff "compare" operation on two files results in the message "Invalid Expression" being printed on stderr followed by an abnormal termination. I've gotten far enough in debugging to determine that this happens when the string "?/entity" is passed to xmlXPathCompile, but I don't understand XML::Diff well enough to know where that is coming from. That string does not appear in our XML input (which appears to be correct and valid). Only some input files trigger the problem, and even files with fairly similar size or contents may not show the problem. It feels a little like a memory corruption or resource error, but the same input triggers the problem on a variety of different systems (Solaris 2.5.1 sparc, Solaris 8 sparc, Solaris 8 intel, RedHat Linux 7.3, RedHat Linux 9, and Fedora 8). Here's how to reproduce it. First, run this shell script to generate the files "string1.xml" and "string2.xml" ############################################################### COUNT=${1:-700} echo '<entitytype n="SSSSSSSSSSSSS">' > string1.xml for i in `seq $COUNT` do cat >> string1.xml <<! <entity> <keyfld n="AAAAAAA" v="1" /> <keyfld n="BBBBBBBBBB" v="28" /> <keyfld n="CCCCCCCCCC" v="175" /> <keyfld n="DDDDDDDD" v="$i-1" /> <keyfld n="EEEEEEEEE" v="1" /> <fld n="TEST" v="$i-1" /> <fld n="FFFFFFFFFFF" v="No" /> <fld n="GGGGGGGGGGGGGGGGGGGGGGGGGGG" v="20" /> <fld n="HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH" v="0" /> <fld n="IIIIIIIIIIIIIIIIIIIIIIIII" v="3.0" /> <fld n="JJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJ" v="5.0" /> <fld n="KKKKKKKKKKKKKKKKKKKKK" v="-7.0" /> <fld n="LLLLLLLLLLLLL" v="Unspecified" /> <fld n="MMMMMMMMMMMMMM" v="No" /> <fld n="NNNNNNNNNNNNNNNNNNNNNNNNNN" v="_" /> <fld n="OOOOOOOOOOOOOOOOOOOOO" v="0" /> <fld n="PPPPPPPPPPPPPPPPP" v="59" /> <fld n="QQQQQQQQQQQQQQQQQQQQQ" v="0" /> <fld n="RRRRRRRRRRRRRRRRRRRR" v="0" /> <fld n="SSSSSSSSSSSSSSSSSSSSS" v="0" /> <fld n="TTTTTTTTTTTTTTTTTTTTT" v="0" /> <fld n="UUUUUUUUUUUUUUUUUUUUU" v="0" /> <fld n="VVVVVVVVVVVVVVVVVVVVV" v="0" /> <fld n="WWWWWWWWWWWWWWWW" v="Unspecified" /> <fld n="XXXXXXXXXXXXXXXXXXXXXXXXXXX" v="Unspecified" /> <fld n="ZZZZZZZZZZZZZZZZZZZZZZZ" v="_" /> <fld n="aaaaaaaaaaaaaaaaaaaaaaaaa" v="Unspecified" /> <fld n="bbbbbbbbbbbbbbbbbbb" v="_" /> <fld n="cccccccccccccccccccc" v="_" /> <fld n="dddddddddddddddddd" v="_" /> <fld n="eeeeeeeeeeeeeeeeeeee" v="_" /> <fld n="ffffffffffffffffffffffffff" v="_" /> <fld n="ggggggggggggggggggggggggg" v="_" /> <fld n="hhhhhhhhhhhhhh" v="_" /> <fld n="iiiiiiiiiiiiiiiiiii" v="_" /> <fld n="jjjjjjjjjjjjjjjjjjjjj" v="_" /> <fld n="kkkkkkkkkkkkkkkkkk" v="_" /> <fld n="lllllllllllllllll" v="_" /> <fld n="mmmmmmmmmmmmmmmmmm" v="_" /> <fld n="nnnnnnnnnnnnnnnnnn" v="_" /> <fld n="oooooooooooooooooooo" v="_" /> <fld n="pppppppppppppppppppppppppppp" v="_" /> <fld n="rrrrrrrrrrrrrrrrrrrrrrrrrrr" v="_" /> <fld n="ssssssssssssssssssssss" v="_" /> <fld n="tttttttttttttttttttttttttttt" v="Disabled" /> <fld n="uuuuuuuuuuuuuuuuuuuuuuuu" v="0" /> <fld n="vvvvvvvvvvvvvvvvvvvvvv" v="0" /> </entity> ! done echo '</entitytype>' >> string1.xml echo '<entitytype n="SSSSSSSSSSSSS">' > string2.xml for i in `seq $COUNT` do cat >> string2.xml <<! <entity> <keyfld n="AAAAAAA" v="1" /> <keyfld n="BBBBBBBBBB" v="28" /> <keyfld n="CCCCCCCCCC" v="175" /> <keyfld n="DDDDDDDD" v="$i-1" /> <keyfld n="EEEEEEEEE" v="1" /> <fld n="test" v="$i-2" /> <fld n="FFFFFFFFFFF" v="No" /> <fld n="GGGGGGGGGGGGGGGGGGGGGGGGGGG" v="20" /> <fld n="HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH" v="0" /> <fld n="IIIIIIIIIIIIIIIIIIIIIIIII" v="3.0" /> <fld n="JJJJJJJJJJJJJJJJJJJJJJJJJJJJJJJ" v="5.0" /> <fld n="KKKKKKKKKKKKKKKKKKKKK" v="-7.0" /> <fld n="LLLLLLLLLLLLL" v="Unspecified" /> <fld n="MMMMMMMMMMMMMM" v="No" /> <fld n="NNNNNNNNNNNNNNNNNNNNNNNNNN" v="_" /> <fld n="OOOOOOOOOOOOOOOOOOOOO" v="0" /> <fld n="PPPPPPPPPPPPPPPPP" v="59" /> <fld n="QQQQQQQQQQQQQQQQQQQQQ" v="0" /> <fld n="RRRRRRRRRRRRRRRRRRRR" v="0" /> <fld n="SSSSSSSSSSSSSSSSSSSSS" v="0" /> <fld n="TTTTTTTTTTTTTTTTTTTTT" v="0" /> <fld n="UUUUUUUUUUUUUUUUUUUUU" v="0" /> <fld n="VVVVVVVVVVVVVVVVVVVVV" v="0" /> <fld n="WWWWWWWWWWWWWWWW" v="Unspecified" /> <fld n="XXXXXXXXXXXXXXXXXXXXXXXXXXX" v="Unspecified" /> <fld n="ZZZZZZZZZZZZZZZZZZZZZZZ" v="_" /> <fld n="aaaaaaaaaaaaaaaaaaaaaaaaa" v="Unspecified" /> <fld n="bbbbbbbbbbbbbbbbbbb" v="_" /> <fld n="cccccccccccccccccccc" v="_" /> <fld n="dddddddddddddddddd" v="_" /> <fld n="eeeeeeeeeeeeeeeeeeee" v="_" /> <fld n="ffffffffffffffffffffffffff" v="_" /> <fld n="ggggggggggggggggggggggggg" v="_" /> <fld n="hhhhhhhhhhhhhh" v="_" /> <fld n="iiiiiiiiiiiiiiiiiii" v="_" /> <fld n="jjjjjjjjjjjjjjjjjjjjj" v="_" /> <fld n="kkkkkkkkkkkkkkkkkk" v="_" /> <fld n="lllllllllllllllll" v="_" /> <fld n="mmmmmmmmmmmmmmmmmm" v="_" /> <fld n="nnnnnnnnnnnnnnnnnn" v="_" /> <fld n="oooooooooooooooooooo" v="_" /> <fld n="pppppppppppppppppppppppppppp" v="_" /> <fld n="rrrrrrrrrrrrrrrrrrrrrrrrrrr" v="_" /> <fld n="ssssssssssssssssssssss" v="_" /> <fld n="tttttttttttttttttttttttttttt" v="Disabled" /> <fld n="uuuuuuuuuuuuuuuuuuuuuuuu" v="0" /> <fld n="vvvvvvvvvvvvvvvvvvvvvv" v="0" /> <fld n="wwwwwwwwwwwwwwwwwwwwwwwww" v="_" /> <fld n="xxxxxxxxxxxxxxxxxxxxxxxxxxx" v="_" /> <fld n="zzzzzzzzzzzzzzzzzzzzzz" v="_" /> <fld n="111111111111111111111111" v="_" /> <fld n="2222222222222222" v="_" /> <fld n="3333333333333333333" v="_" /> </entity> ! done echo '</entitytype>' >> string2.xml ############################################################### Then, run this perl program. (It assumes that XML::Diff is installed) The expected result is normal termination. Instead, the message "Invalid expression" is printed and the program exits with an error. ############################################################### #!/usr/bin/perl use XML::Diff; my $diff = XML::Diff->new(); my $diffgram = $diff->compare( -old => "string1.xml", -new => "string2.xml" ); ############################################################### Thanks! -- Peter Fales Alcatel-Lucent Member of Technical Staff 2000 Lucent Lane Room: 1C-436 Naperville, IL 60566-7033 Email: psfales@alcatel-lucent.com Phone: 630 979 8031