RPM creation questions

Why should there be a file INSTALL inside the rpm, and why should it contain any commands?

Have a look at some existing source rpm files, install them, read the spec file, and figure out how it works. Don't start with a complicated thing like a kernel, rather look at a simple one like cflow.

Then google for "maximum rpm". I never remember the exact name of the document, but that is the best documentation of rpm I am aware of, though it is perhaps a bit outdated.

Consider using a %postinstall section in the spec file if you need to do more than just copying files into the user's system, in the binary rpm.

Source rpm's just install the tarballs and patches inusr-src-something-distro-specific-SOURCES, and the spec file inusr-src-something-distro-specific-SPECS.

??? In the home directory ???

The install step is supposed to build a copy of the file system that the user will see when the binary rpm is installed. This is built under $RPMBUILDROOT, which usually points tovar-tmp-%{name}-%{version}-%{release} or something similar. The subsequent %files section is used to pick files from this place and package them in the binary rpm.

Notice that the prep, build, and install sections are used during the "rpmbuild -ba" command, and their commands are not used when the binary rpm is installed.

-Enrique