running "make clean" in xfsprogs doesn't wipe mkfs/*.o files
I recently ran into an issue where "make clean" (or "make realclean") in the xfsprogs sources isn't deleting the *.o files in the mkfs directory.
This is causing a problem because I first build xfsprogs for arm64, and then I build it for x86_64, with a "make realclean" between. The fact that object files in the mkfs directory aren't getting cleaned results in build failures. This wasn't happening before when I was building on a different build machime, but it's pretty clear the problem is for soem reasone "make clean" isn't deleting the object files in mkfs.
I tried to debug xfsprogs's makefiles, but it wasn't immediately obvious, so I thought I would just file a report and hope someone can take a look at it.
In the meantime, I'll work around it by adding a "find . -type f -name \*.o -print0 | xargs -0 rm" to my test appiance build script. Which should work, but it seems really unfortunate that it's necessary.
Thanks,
- Ted
I don't have an answer either, and haven't seen this one in particular, but the xfsprogs build system is special at best or broken if you want to be frank. A while ago I started on a ninja + meson port as that seems to be build system of choice for many low-level tools. I didn't really have time to do more than a minimal demo, buf if someone wants to move to a less broken build system I'd be 100% supportive.
Quoted 6 lines · Christoph Hellwig
I don't have an answer either, and haven't seen this one in particular, but the xfsprogs build system is special at best or broken if you want to be frank. A while ago I started on a ninja + meson port as that seems to be build system of choice for many low-level tools. I didn't really have time to do more than a minimal demo, buf if someone wants to move to a less broken build system I'd be 100% supportive.
Yeah, ninja+meson would be nice, whole probing system would be much simpler.
Signature · 1 line
- Andrey