On 7/22/26 11:41 AM, Jeremy Bícha wrote:
It looks like the blhc failure is simply because the build steps done by meson install are not verbose enough for blhc's analysis of the build logs. See https://salsa.debian.org/gnome-team/extras/quadrapassel/-/jobs/9791972 with this build log excerpt 3239:NONVERBOSE BUILD: [3/10] Compiling C object src/quadrapassel.p/meson-generated_.._resources.c.o I don't know why meson install insists on rebuilding some things during the install step. Nevertheless, for Debian's purposes, I believe we can rely on things being built during meson's build step and avoid an unnecessary rebuild (for our purposes) during meson install.
Meson specifically allows you to have steps that depend on the install target but not the "all" target. These are typically for things that would be slow and unnecessary during development but needed for system installs. The problem here is not so much that some build steps happen during install, but rather that "ninja install" is not using "-v". If it did, the build steps would be shown and blch would be (presumably) happy.
If there are spurious rebuilds for "ninja install" after a "ninja all (or test"), then that is either a bug in Meson or in the project's build files. This can be a custom script that touches its input files or what have you.
