Red Hat Bugzilla – Bug 458577
Change default fuzz to something more sensible
Last modified: 2008-08-12 01:19:08 EDT
Description of problem: Zero fuzz when applying patches makes many routine package maintenance task way more difficult. Rediffing the patch while retaining comments takes time, even if equipped with patchutils. Our CVS make tooling tries to mitigate that with rediff and patch targets, but they're useless in case your source explodes to something different than %{name}-%{version} or you change a single file with multiple patches. They could surely be fixed/enhanced, but I'm not quite sure rpm should rely on that. Therefore, please restore the default to 0. I am very sure that number of packages that benefit from zero fuzz is much smaller than ones, that benefit from patch guessing the right place to apply the fuzz, and not rediffing the patch and shuffiling the files around in CVS (due to file name change).
You can always use fallback to the original fuzz like: %patch1 -F3 -p1 -b .backup so you don't need to rediff patches, simply add -F3 when you are sure what you are doing. The fuzz=0 is the only acceptable default for applying patches from least-surprise POV. Please try to live with it. It pretty much makes sense.
Or stick "%define _default_patch_fuzz 2" (2, not 3 is what patch defaults to wrt fuzz) to your specs to avoid having to put it to each %patch command separately.