Re: [PATCH] Fix PR94401 by considering reverse overrun

[Jakub Jelinek <jakub@redhat.com>]

On Thu, Apr 02, 2020 at 06:07:55PM +0800, Kewen.Lin wrote:
> > The above is misformatted.  The ? and : shouldn't be indented further than
> > the dataref_offset, but usually e.g. for the sake of emacs we add ()s around
> > the expression in this case.  So:
> > 			tree offset = (dataref_offset
> > 				       ? dataref_offset
> > 				       : build_int_cst (ref_type, 0));
> > or
> > 			tree offset
> > 			  = (dataref_offset
> > 			     ? dataref_offset : build_int_cst (ref_type, 0));
> > 
> 
> Thanks Jakub!  I'll follow this by add () for ternary expression.
> With manual added "()", clang-format can get below:

Note, the () isn't about ternary expressions, if everything fits on one
line, there is no reason to add ()s, so
  tree offset = dataref_offset ? dataref_offset : build_int_cst (ref_type, 0);
is just fine that way, on the other side
					int whatever = HOST_WIDE_INT_1U
						       + foobarbaz (qux);
should have them too, like:
					int whatever = (HOST_WIDE_INT_1U
							+ foobarbaz (qux));
or
					int whatever
					  = HOST_WIDE_INT_1U + foobarbaz (qux);
I don't use emacs, but was told that emacs without the ()s would misindent
it like (I think):
					int whatever = HOST_WIDE_INT_1U
					  + foobarbaz (qux);
which is what we do not want.

> 
> 			tree offset
> 			  = (dataref_offset ? dataref_offset
> 					    : build_int_cst (ref_type, 0));
> 
> contrib/check_GNU_style.sh didn't complain this, I'm not sure whether
> it's possible to add this kind of convention into contrib/clang-format.

clang-format is not our official indentation style; I have no problem with
the above formatting from readability POV, though unsure what emacs will do
with that (but if it moves that : right below the first dataref_offset,
no big deal, that is also fine and probably more appropriate if the
build_int_cst... is long and would need more wrapping).

	Jakub