Regarding METRON-1087, I’m in favor of freezing commits for a day, to let
Justin re-run the script for METRON-1087 over all of current master, and commit
it.
Perhaps, for assurance, this commit should only include the fully-automated
“vast majority”; the couple dozen files that needed manual fixes could be split
out into a patch that gets manual review. I don’t have a problem saying +1 on
what is evidently script-automated (L1:’s#/**#/*#’ and blank line after license
block).
All of us with open PRs will then have to update to master, but we do that
periodically anyway. Github is really quite good about not causing extra
review work for update merges.
IMHO,
--Matt
On 8/11/17, 5:14 AM, "Justin Leet" <justinjl...@gmail.com> wrote:
Now that METRON-746 <https://github.com/apache/metron/pull/577> is in, we
have a consistent code formatting setup where (for the most part) it can be
autoapplied.
Barring one existing PR, the main thing is just picking a module, doing the
format, and testing it out. Obviously, I'd like to avoid collisions with
any major PRs out there (say METRON-777
<https://github.com/apache/metron/pull/530>), so things like parsers are
out.
Does anybody have any thoughts on what should be grabbed first? Maybe
metron-stellar since it's pretty easy to test and even though it gets PRs,
they're typically fairly small and contained?
The main hiccup before being able to do any blanket reformatting is this PR:
METRON-1087: Adjust license headers to be comments instead of Javadoc
<https://github.com/apache/metron/pull/685>
Without it, all the license headers get reformatted in Javadoc style when
autoformatted (this actually happened before, but since we didn't actually
format our code much it was pretty benign). Once this is in, I'm fine doing
any headers that sneak in via PRs, it's a pretty easy find/replace in the
editor.
Once we're confident this works fairly smoothly, it should be pretty easy
to branch out and start hitting more modules in whatever priority order we
want.
Justin
