On 3/12/07, Alfred von Campe <alfred at 110.net> wrote: > On Mar 12, 2007, at 10:15, Jim Perrin wrote: > > >> 3. After copying the config file an making one small > >> change with "make xconfig", why are there so many > >> differences in the config file? > > > > Are they actual changes, or is the diff now simply off by a few lines? > > I went back and started the process of building a kernel from scratch > on a different system. This time I paid closer attention to what was > going on. After the "rpmbuild -bp --target=`uname -m` > kernel-2.6.spec" step, the ~/rpmbuild/BUILD/kernel-2.6.9/linux-2.6.9 > directory contains among other things these two files: > > -rw-r--r-- 1 rpmbuild rpmbuild 50332 Mar 12 15:32 .config > -rw-r--r-- 1 rpmbuild rpmbuild 61600 Mar 12 15:31 .config.old > > Notice the size difference of these two files. After copying the > config file from the ~/rpmbuild/SOURCES directory, these two files > now look like this: > > -rw-rw-r-- 1 rpmbuild rpmbuild 61600 Mar 12 15:42 .config > -rw-r--r-- 1 rpmbuild rpmbuild 61600 Mar 12 15:35 .config.old > > And a diff confirms the the two files are now identical. However, > after doing a "make xconfig", changing two settings in the Serial > Port section and saving the config file, the directory looks like this: > > -rw-rw-r-- 1 rpmbuild rpmbuild 50322 Mar 12 15:44 .config > -rw-rw-r-- 1 rpmbuild rpmbuild 61600 Mar 12 15:42 .config.old > > Finally, comparing the new .config file with the original .config > file only shows the expected changes: > > # diff /tmp/saved_configs/.config .config > 4c4 > < # Mon Mar 12 15:32:01 2007 > --- > > # Mon Mar 12 15:44:32 2007 > 1501c1501 > < CONFIG_SERIAL_8250_NR_UARTS=4 > --- > > CONFIG_SERIAL_8250_NR_UARTS=16 > 1503c1503 > < # CONFIG_SERIAL_8250_MANY_PORTS is not set > --- > > CONFIG_SERIAL_8250_MANY_PORTS=y > > While I don't quite understand what is going on, it appears that > xconfig is doing the right thing. I'm rebuilding the kernel again as > I type this. BTW, there are lots of compiler warnings: is this > normal/expected for a kernel build? > Its kind of a sad state of affairs but the kernel since as far back as I can remember has been chock full of warnings, such that the warnings are useless to someone like you and me. Only the developers of different parts can tell which are meaningful are not...james