I have been looking into the issues with the insserv test scripts. There are a few problems here, in brief:
1. The test scripts were looking in the old area for insserv output and missing some dependency information that way. I should have updated them earlier. (This has been fixed.) 2. The scripts were not well integrated into the Makefile process. That is to say running "make check" and "make distclean" should have been handling the scripts better. (This has also been fixed.) 3. The scripts were causing insserv to try to overwrite the host system's start-up dependency information in some instances, which could cause problems if run as root/sudo. (This seems unlikely to happen, but I've fixed this too.) 4. Some of the tests are expecting output which is different than what insserv provides and, possibly, should provide. For instance, some scripted tests fail if scripts without LSB headers are not inserted into runlevels in a certain order. This one seems strange to me because, at the moment, insserv does not try to place non-LSB scripts in runlevels in any order, considering it something to be skipped over. In other words, I think there is some difference in expectations between what I think insserv should be doing and what the scripts we inherited from downstream think insserv should be doing. I'm interested in getting some feedback on this. If a script has no LSB header, should it be given preset defaults, or should it be ignored? If the latter, I think we can remove this series of tests. Which will mean there are just two errors left for me to sort out. - Jesse
