Re: [boost] Policy/ethics question: correcting 'trivial' errors in others' code?

From: Beman Dawes (bdawes_at_[hidden])
Date: 2002-01-16 15:45:58

At 03:09 PM 1/16/2002, David Abrahams wrote:
 >>   While looking at some of the regression testing failures on the
 >dashboard
 >> (see my earlier post), I've noticed that many errors are trivial to 
fix.
 >> Should we fix the errors in CVS immediately or construct a patch, 
notify
 >the
 >> author, and wait for approval? What if the author does not respond in a
 >> timely manner? Some of these problems really are trivial, e.g., look at
 >the
 >> failures for the "compose" library, where there is a #include
 >"compose.hpp"
 >> that should be #include <boost/compose.hpp>.
 >>
 >>   Any guidelines on this?
 >
 >The usual practice is that editorial errors like the above can be fixed
 >without consulting anyone, but that you should err on the side of asking
 >the
 >author. One thing you can do is to check in a branch and ask the author 
to
 >approve it before you merge.

I suppose we ought to write up the guidelines.

One thing I've done is to go ahead and make the change in my working copy 
without committing it, and notify the developer.  Because it lights up on 
WinCvs, I don't forget the file is changed.  If the developer doesn't 
respond in a reasonable time, then I go ahead and commit the change.

The details aren't important; the objective is to mostly let the developer 
do the fixes, but not let fixes languish for a long time because the 
developer isn't responding.

--Beman