I'd love to have a standardized, conditional test section of the specfile - it would be useful for teststuites. Current suggestion is to "add a %test section in the specfile, corresponding --runtest commandline option. Nonzero exit == failure?"/
Comments from jakub: The question is what this %test script can expect in which variables from rpm and what directory will it be changed into before starting. My main question is whether its builddir will be unpacked and in what variable will get it, and whether its instdir will be populated or not (and again, where to query that). For the packages I care about, I only need builddir to be unpacked and after at least -bc phase.
There are several ideas here that need to be unsnarled so that I can attempt an implementation. BTW, there's an equivalent need for an install, not a build, scriptlet to verify that the bits are, indeed, installed correctly. That might simplify many QA and support issues if attempted. The name %test is woefully underdescriptive <shrug>. Where in the build state machine do you want %test executed? Under what conditions? What sort of CLI gunk do you want to execute the script? Jakub: the cwd/environment for all scriptlets is deterministic, can even be reconfigured using %__spec_test_stage and/or %___build_pre if/when desired. Yes this is more mechanism than policy, and makes everyones head hurt a lot, but is sufficiently general to handle a new %test section quite easily and cheaply.
Guys, I need a definition owf what is being requested.
Closing bugs on older, no longer supported, releases. Apologies for any lack of response. For RPM issues, please try a current release such as Fedora Core 4; if bugs persist, please open a new issue.