svn commit: r345707 - in head: lib/clang lib/libc++ lib/libc++experimental lib/libc++fs lib/libc/tests/stdlib lib/libclang_rt lib/libcxxrt lib/libgcc_eh lib/libomp lib/ofed/libibnetdisc share/mk us...

[Warner Losh]

On Fri, Mar 29, 2019, 9:15 PM Enji Cooper <yaneurabeya at gmail.com> wrote:

>
> > On Mar 29, 2019, at 8:09 PM, Rodney W. Grimes <freebsd at gndrsh.dnsmgr.net>
> wrote:
> >
> >> Author: ngie
> >> Date: Fri Mar 29 18:43:46 2019
> >> New Revision: 345707
> >> URL: https://svnweb.freebsd.org/changeset/base/345707
> >>
> >> Log:
> >>  Revert r345706: the third time will be the charm
> >>
> >>  When a review is closed via Phabricator it updates the patch attached
> to the
> >>  review. I downloaded the raw patch from Phabricator, applied it, and
> repeated
> >>  my mistake from r345704 by accident mixing content from D19732 and
> D19738.
> >
> > Which, arguable is a feature or mis feature depending on the point
> > of view.  I do not like it when I go to look at someone elses
> > committed code siting a review, as I want to actually see what
> > it was that was committed.  You can find the pre-commit diff,
> > but it takes a bit of probling.  The upside is you can get
> > both diffs from the same place and diff the diffs :-)
> >
> >>  For my own personal sanity, I will try not to mix reviews like this in
> the
> >>  future.
> >
> > :-)  Been there, almost did that too.
> > Pre commit last minute svn diff saved me.
>
> …
>
> This is why I’m doing the following from here on out:
>
> $ arc patch
> $ svn ci
>
> Unfortunately svn doesn’t support all of the niceties of “arc land”.
> Otherwise, I would have used that.
>
> The Facebook version of “arc land” (before their new non-public variation)
> supported verifying diffs in local repos vs Phabricator to make sure that
> the diff content was consistent/correct.
>
> * Pro: it would catch issues like what I did the first time.
> * Con: I couldn’t make last minute changes (I would need to resubmit the
> change and have it re-reviewed, which I argue is a good feature).
>
> Just some food for thought. For now, arc patch/svn ci works for me.
>

I use git svn and git show every commit so I'm sure when I want to do
multiple at once. Phab doesn't make that easy, alas, so I often choose the
lesser evil of just getting a review on the squashed branch. I only split
when there is a good reason. If we had better tools, I wouldn't be forced
to pick, but you go to wat with the tools you have...

Warner

>