On 05/21/2015 04:49 PM, Dale wrote: > Mike Gilbert wrote: >> On Thu, May 21, 2015 at 3:44 PM, <[email protected]> wrote: >>> walt <[email protected]> wrote: >>> >>> [...] >>>> Then, after I figured out that CONFIG_USER_NS is a kernel config item, >>>> requiring reinstallation of my kernel, I wasted more time figuring out >>>> (for the n'th time) that you shouldn't just change a single kernel >>>> config item and do "make" because that shortcut can break important >>>> things. >>>> >>>> No, you should do "make clean" first, and then do "make" etc. >>> [...] >>> >>> I haven't done a "make clean" for years when I compiled a kernel and I >>> never had any problems. >> Then you have not made any critical config changes, or you have been very >> lucky. >> >> > > Then so have I. I have changed one thing a lot of times over the years, > run make and it work fine. Most of the time, it is when emerge spits > out that a option is needed for a package to work. Honestly, this is > the first time I recall hearing this should even be done.
The first n times I discovered than "make clean" prevents (some) problems (sometimes) is when I was running the daily unstable kernels directly from Linus's git repo. As you would expect, I had to git-bisect a lot of kernel bugs over the years, and along the way I discovered that doing the exact-same bisect on the exact- same source code could produce different results -- results that were just plain wrong sometimes. That problem disappeared when I started doing "make clean" after every bisect, painful though it seemed at the time. I'm now much too old and grouchy to debug unstable kernels every day, though.
