Patches with independent changes

[Paul Eggert <eggert@cs.ucla.edu>]

[retitling from "Changes in update-game-score.c"]

Eli Zaretskii wrote:

> chmod is documented to be able to return both negative and positive values

Really?  Where is it documented to do that?  It's not documented that 
way in the POSIX documentation for chmod, or in the Gnulib documentation 
of chmod porting hassles; see 
<http://pubs.opengroup.org/onlinepubs/9699919799/functions/chmod.html> 
and <http://www.gnu.org/software/gnulib/manual/html_node/chmod.html>, 
respectively.

Even if you're correct, which I doubt, trunk bzr 116116 would still be 
an example of what you're calling a "sumo" patch -- a patch containing 
multiple independent changes. One change addresses WINDOWSNT's lack of 
fchmod. Another addresses the perhaps-hypothetical issue that chmod can 
return positive values. So regardless, that patch does not follow the 
rule that patches should fix just one bug and should not make 
independent changes.

Insisting on such a rule would be silly, anyway. We don't follow that 
rule and we never have, and we shouldn't impose it now. Obviously there 
are advantages to putting independent changes in separate patches and in 
many cases we should do that, but an extremist insistence of this 
principle in all cases would significantly increase our maintenance 
burden for little or no benefit.