tree 7e79b62722b87cd25d3c61933e1fc09d22daac2b
parent d874bb911222d211f0201ae1cd4b73428e7c1030
author Scott Graham <scottmg@chromium.org> 1579637683 -0800
committer Commit Bot <commit-bot@chromium.org> 1579731303 +0000

format: Fix not getting to fixed point in one step

In https://gn-review.googlesource.com/c/gn/+/7140/ some styles of suffix
comments the formatter would wrap the suffix comments. But on reapplying
the formatter a second time, the parse tree is slightly different. The
second time through, the wrapped comment is the only thing on a line
separating two statements.

  a =                 1
    b + c  # x        2
           # y        3
  d = e               4

The ParseNode doesn't have a LocationRange that includes the suffix
comments, so when the formatter attempts to determine if there was a
blank line between two statements, it incorrectly decides there should
be an additional blank line added (because the end of the first
statement is not directly next to the start of the next one). That is,
the "end" of the "a=b+c" is line 2, not line 3.

I fiddled around with trying to amend the parse tree to make the
comments be included in the range of the node, but I didn't come up with
a reasonable way to do that. Instead, have the formatter calculate the
"real" range at formatting time, so instead of simply using line numbers
for A and B to determine if they're split, it walks down A looking for
suffix comments and finds the lowest down one and uses that instead. In
the example above, the walk will determine that "a=b+c" really ends on
line 3, so is adjacent to "d=e" on line 4 and no blank line should be
inserted.

This makes the recent Chromium test reformat
https://chromium-review.googlesource.com/c/chromium/src/+/2011169/
"stick" in that no further changes are made if the formatter is run on
it again.

Bug: gn:141
Change-Id: If320a87feec9671d2368a799e9e396866698561d
Reviewed-on: https://gn-review.googlesource.com/c/gn/+/7180
Commit-Queue: Scott Graham <scottmg@chromium.org>
Reviewed-by: Brett Wilson <brettw@chromium.org>
Reviewed-by: Nico Weber <thakis@chromium.org>