The situation is a little confusing. The respective unit test only fails when it runs as part of 'make check'. When executed individually it runs and passes fine.
Moreover, 'make check' passes when the package is built without optimization (i.e. -O0) That makes it a little difficult to debug. Valgrind reports: ==28928== 1 errors in context 1 of 60: ==28928== Conditional jump or move depends on uninitialised value(s) ==28928== at 0x494ABA0: Lexer::tokenizePunctOperator() (lexer.cpp:579) ==28928== by 0x4949F90: Lexer::PeekToken(Lexer::TokenValue*) (lexer.cpp:254) ==28928== by 0x495AC92: ClassAdParser::parseClassAd(ClassAd&, bool) (source.cpp:1161) ==28928== by 0x4957D15: ClassAdParser::ParseClassAd(LexerSource*, ClassAd&, bool) (source.cpp:211) ==28928== by 0x4957C97: ClassAdParser::ParseClassAd(std::istream&, ClassAd&, bool) (source.cpp:199) ==28928== by 0x804A928: read_from_stream_alt(ClassAd**, ClassAd**) (extra_tests.cpp:344) ==28928== by 0x8049E44: test_parsing() (extra_tests.cpp:135) ==28928== by 0x8049C33: main (extra_tests.cpp:89) ==28928== Uninitialised value was created by a stack allocation ==28928== at 0x494B56F: InputStreamLexerSource::ReadCharacter() (lexerSource.cpp:111) which might have something to do with the problem. ... -- Michael Hanke http://mih.voxindeserto.de:q -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org