Re: [PATCH v2] docs: checkpatch: add UNNECESSARY_ELSE message

[Lukas Bulwahn]

On Sun, Oct 3, 2021 at 6:38 AM Dwaipayan Ray <dwaipayanray1@xxxxxxxxx> wrote:
>
> On Sat, Oct 2, 2021 at 8:15 PM Utkarsh Verma <utkarshverma294@xxxxxxxxx> wrote:
> >
> > Added and documented UNNECESSARY_ELSE message type.
> >
> > Signed-off-by: Utkarsh Verma <utkarshverma294@xxxxxxxxx>
> > ---
> > Changes in v2:
> >   - Included the continue statement.
> >
> >  Documentation/dev-tools/checkpatch.rst | 77 ++++++++++++++++++++++++++
> >  1 file changed, 77 insertions(+)
> >
> > diff --git a/Documentation/dev-tools/checkpatch.rst b/Documentation/dev-tools/checkpatch.rst
> > index f0956e9ea2d8..b7c41e876d1d 100644
> > --- a/Documentation/dev-tools/checkpatch.rst
> > +++ b/Documentation/dev-tools/checkpatch.rst
> > @@ -1166,3 +1166,80 @@ Others
> >
> >    **TYPO_SPELLING**
> >      Some words may have been misspelled.  Consider reviewing them.
> > +
> > +  **UNNECESSARY_ELSE**
> > +    Using an else statement just after a return/break/continue statement is
> > +    unnecessary. For example::
> > +
> > +      for (i = 0; i < 100; i++) {
> > +              int foo = bar();
> > +              if (foo < 1)
> > +                      break;
> > +              else
> > +                      usleep(1);
> > +      }
> > +
> > +    is generally better written as::
> > +
> > +      for (i = 0; i < 100; i++) {
> > +              int foo = bar();
> > +              if (foo < 1)
> > +                      break;
> > +              usleep(1);
> > +      }
> > +
> > +    It helps to reduce the indentation and removes the unnecessary else
> > +    statement. But note, there can be some false positives because of the
> > +    way it is implemented in the checkpatch script. The checkpatch script
> > +    throws this warning message if it finds an else statement and the line
> > +    above it is a break/continue/return statement indented at one tab more
> > +    than the else statement. So there can be some false positives like::
> > +
> > +      int n = 15;
> > +      if (n > 10)
> > +              n--;
> > +      else if (n == 10)
> > +              return 0;
> > +      else
> > +              n++;
> > +
> > +    Now the checkpatch will give a warning for the use of else after return
> > +    statement. If the else statement is removed then::
> > +
> > +      int n = 15;
> > +      if (n > 10)
> > +              n--;
> > +      else if (n == 10)
> > +              return 0;
> > +      n++;
> > +
> > +    Now both the n-- and n++ statements will be executed which is different
> > +    from the logic in the first case. As the if block doesn't have a return
> > +    statement, so removing the else statement is wrong.
> > +
> > +    Always check the previous if/else if blocks, for break/continue/return
> > +    statements, and do not blindly follow the checkpatch advice. One
> > +    patch (https://lore.kernel.org/all/20200615155131.GA4563@sevic69/)
> > +    even made it to the mainline, which was again reverted and fixed.
> > +    Commit 98fe05e21a6e ("staging: rtl8712: Remove unnecesary else
>
> s/unnecesary/unnecessary
> > +    after return statement.")
> > +
> > +    Also, do not change the code if there is only a single return statement
> > +    inside if-else block, like::
> > +
> > +      if (a > b)
> > +              return a;
> > +      else
> > +              return b;
> > +
> > +    now if the else statement is removed::
> > +
> > +      if (a > b)
> > +              return a;
> > +      return b;
> > +
> > +    there is no considerable increase in the readability and one can argue
> > +    that the first form is more readable because of the indentation. So
> > +    do not remove the else statement in case of a single return statement
> > +    inside the if-else block.
> > +    See: https://lore.kernel.org/lkml/20140925032215.GK7996@xxxxxxxxxxxxxxxxxx/
> > --
> > 2.25.1
> >
>
> I think this message is unnecessarily long for a warning that's understandable
> at best without the verbose part. Try to shorten it up with only what's
> required for a user to understand why the warning is there.
>

Dwaipayan, I actually considered all this interesting information and
all valuable background information on this rule.

Now, I would like to see all this information in the checkpatch
documentation. Maybe here, the expectations for the --verbose option
and the checkpatch documentation are slightly different.
IMHO, the need for the checkpatch documentation beats the --verbose
option. If checkpatch users really ask for --verbose help on this
rule, they are already questioning the value of a rule that is already
quite understandable (as you said). So, then we should convince them
with all background information and known false positives we
encountered.

I vote for keeping all information; wordsmithing and writing more
precisely is certainly doable.

Lukas