On Sun, 2006-06-18 at 16:22 -0400, bugzilla@xxxxxxxxxx wrote: > Please do not reply directly to this email. All additional > comments should be made in the comments box of this bug report. > > Summary: Review Request: rasqal - RDF query library > > > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=195645 > > > > > > ------- Additional Comments From thomas@xxxxxxxxxxxxx 2006-06-18 16:14 EST ------- > It's not that I do not want to use %makeinstall - it's that the packaging > guidelines that mention you shouldn't use it do not say *anything* useful about > it, why it's bad, and they're not even correct. If I need to make this change, > I need to know because I have a bunch of other packages using %makeinstall. > > Let's take those rules step by step: > > %makeinstall overrides a set of environment variables during "make > install". I.e. it performs make prefix="..." includedir="..." ... > > This is wrong - it overrides make variables, not environment variables. Unless make isn't broken and unless the makefiles aren't playing dirty games with make flags the effect is the same. > In addition, this is all the rule says - it does not say that it is wrong or why > it is wrong. So it's a) factually wrong and b) irrelevant. > > > It is error prone, and can have unexpected effects when run against less > than perfect Makefiles. > > How is it error prone ? - More sources of errors: A dozen vars vs. one single var. - %makeinstall causes the makefiles to see a different set of variables between "make install" and other (previous/subsequent) invocations of make. > How does make DESTDIR=... not fail when run against a > less than perfect Makefile - for example, one that doesn't even *have* DESTDIR ? > > It can trigger unnecessary rebuilds when executing "make install" > > Don't know about this one, it may be true or may not be true, but in all the > packages I've built I've never known this to be a problem that actually bothered me. It's the second point above. %makeinstall causes the makefiles to see different variables during "make install", than those which had been used in %configure or during "make all". This triggers broken rebuilds, if a makefile contains dependencies on the make variables being changed during %makeinstall. A similar problem occurs with makefile which edit/generate files during "make install" ("install-hooks"), e.g. to propagate final installation dirs to scripts/config-files. > If a package contains libtool archives, it can cause broken *.la files to > be installed. > > I haven't seen broken .la files, This issue for example affects(~ed?) GCC. Ralf
