On Tue, Jul 03, 2018 at 12:29:00PM +0200, Alberto Gonzalez Iniesta wrote: > Have you tested your assertion? Because if ./configure fails, MCONFIG is > not created and the build (make) fails:
That was a general statement. I think it doesn't really matter whether it practically does cause misbuilds: The cost/benefit ratio is fairly obvious. Fixing such issues is generally trivial. Checking whether they really can result in misbuilds is generally expensive. Thus the policy prohibits this bad pattern. The other issue is that delayed failure costs time when diagnosing build failures. You look at the failing step and then notice that some prerequisite was missing/broken until you find the root cause. For instance, a cross build of netkit-tftp fails in dh_strip even though the cause is in the ./configure invocation. That is another reason for not allowing the delaying of failure. The policy not just prohibits silent misbuilds, it goes further and disallows the pattern than can make them happen. Helmut