Re: Patches with independent changes

[Paul Eggert]

Eli Zaretskii wrote:
> otherwise the build would have failed when compiler warnings are
> turned on and used as errors.

That would have been odd, as other static variables in Emacs are 
declared and set but never used (e.g., category_table_version) and 
evidently these builds are not failing.  It's certainly OK to clean up 
the glitches, but such cleanups are independent changes.

> Then any number of patches can be installed in a single commit,
> because they are all "related" -- after all, they are all about Emacs.

No, that's too broad.  Patches should contain changes that are more 
closely-related than that.

> you cannot chmod a file that is open ... the chmod call will fail.

Where is this documented?  Does the problem occur if any process has the 
file open, or only if the current process has it open?  What is errno 
after the failure?  This problem does not occur on POSIXish platforms; 
if it happens under Microsoft Windows the incompatibility should be 
documented (in Gnulib, if the problem is generic to GNU applications).

> the fact that the problem could have been solved in more than
> one way doesn't mean there are multiple changes involved.

True, but if the problem could have been solved in a simpler way that 
involved fewer changes, then multiple changes were most likely present.

Come to think of it, why is that chmod needed at all in WINDOWSNT?  The 
file is already readable and writeable, so as I understand it chmod 644 
is therefore a no-op on that platform.  If so, the attached patch would 
have been simpler yet, no?

update1.diff
Description: Text Data