Olaf,

Can you explain the problem and the effect of the conditional changes and confirm that quilt setup and build will still work with these changes? Pardon my ignorance but I don't follow and quilt setup worked for me months ago without these changes but it might be because I had pre-installed certain packages. Thanks for any input you can provide to improve my picture of the problems and the fix. For now I'm a bit blind to them, probably my own fault.

-- David Mair.

Olaf Hering (olh) - 11 months ago

author

With a "minimal" system with just quilt,vi and osc working on this pkg fails because the macros do not exist. quilt does a rpmbuild -bp *.spec, which expands a the global macros and executes the %prep section. If the macros are missing, the rpmbuild will fail.

---

Petr Tesařík (ptesarik) - 11 months ago

target maintainer

Thanks for the explanation. Enabling "naked" rpmbuild seems reasonable. But IIUC if %suse_kernel_module_package is omitted, then the kernel module is not packaged. Maybe it shouldn't even be built in that case. What about moving the BuildRequires a few lines higher (near kernel-syms) and setting %build_kmp to zero if %suse_kernel_module_package is not defined?

---

Olaf Hering (olh) - 11 months ago

author

All of this does not matter for quilt setup *.spec, which just unpacks the files and applies patches. No build is attempted, no binaries are produced.

I can not draw you a picture, or find any other words to describe the bug, so please experience it yourself:

install a minimal VM, just enough to have network, run zypper, be able to login. sudo zypper in osc sudo zypper in quilt osc co -u openSUSE:Factory/kexec-tools cd $_ osc build standard x86_64 quilt setup *.spec

The build will likely succeed, the quilt command will likely fail because expected packages are missing.

I have to wonder what the kernel macros are useful for if they can simply be removed in a case like this. @ptesarik: Do you know any reason to care that crash is dependent on them. Is removing dependency on them going to have side effects? Why did it ever matter that crash was dependent on them in the first place? ...which is probably a question about years ago activity with no current reference. The signs are that nothing breaks but I'm no convinced I have the whole picture. I'd like to clear this submit (accept or not) and move on as I have one pending patch that's a useful fix for runtime usage of crash and the likelihood of another within days.

Never mind, I read the changes more carefully. I was taking "remove" and "hard" too literally. The change "softens" the kmp build by requiring the dependency kernel_module_package_buildreqs . Requiring it to be defined rather than "assuming" it is defined is the reduction in hardness. It's okay by me. I'll verify today where it sits in the current project since it's been a couple of months.

@dmair, @leonardocf, @ptesarik: review reminder