On Tue, 2022-04-12 at 17:18 +0200, Sébastien Hinderer wrote: > On the PR, David explains the problem in a very detailed way and > proposes a solution. > > Both of us would be interested in a review by GNU make experts.
The discussion of the issue and the proposed solution possibilities look correct to me. An alternative would be to use a "sentinel file" model (where all generated targets depend on a sentinel file, and that sentinel file contains the actual recipe plus a touch command to touch the sentinel file) but I don't think it's needed in this situation. There can still be issues if people go in behind the build system and start deleting and touching files by hand. For example if someone deletes just the target with the empty recipe, make doesn't know how to rebuild it. But otherwise it should be OK.
