On Sat, 06 Jan 2018 at 03:01:35 +0100, Axel Beckert wrote: > Axel Beckert wrote: > > I don't know if that test suite failure shows that expat broke libnih > > or if the test suite just needs to be adapted to the new expat > > version. > > > > - TEST_FILE_EQ (output, ("test:foo:2:0: " > > + TEST_FILE_EQ (output, ("test:foo:1:36: " > > "Invalid object path in <node> name > > attribute\n")); > > TEST_FILE_END (output); > > TEST_FILE_RESET (output);
It certainly seems more true to say that the error in "<node name=\"com/netsplit/Nih/Test\"/>\n" is at line 1 column 36 (the closing double quote around the invalid object path) than at line 2 column 0 (after the newline). > I'd appreciate some more eyes on that change as I have not much of an > idea of libnih's guts. I have no idea about how libnih works, but the change looks harmless at worst - it's not as if it changes the behaviour of non-test code, and the new output looks more correct than the old. (My only interest in libnih is in keeping libpam-systemd installable on buildds.) smcv