Commit message will show up in the Changes dropdown. These are specific to the current range of revisions, but not the current file.
How can you have mixed bounds for "base->r1"
It actually looks like in commit-by-commit it's just
r2?
If that's what got selected automatically then there's a bug in the bounds picking algorithm. :(
And there might me a bug in the algorithm that tries to figure out what the "global" diff bounds are as well.
Basically if I toggle between these two buttons it keeps advancing to
(rN+1,rN+1)
"these two" == "mark as reviewed" <-> "next unreviewed diffs"
Looks like it does, but this keeps me from seeing the commit desc?
So maybe it only happens for files that still exist later?
Looks like it working
Nice, thanks
We'll be starting another review cycle soon
I don't know if it's just the size of the review or if there is actually a more fundamental problem, but some interactions seem to crash chrome. https://reviewable.io/reviews/batfish/batfish/5129#-
@pkaminski
Bug filed: https://bugzilla.mozilla.org/show_bug.cgi?id=1591579
yay looks like this should be fixed in Firefox 72, which goes to beta next week: https://wiki.mozilla.org/Releases/Firefox_72
If the latter, wondering if bug reports / enhancement requests are useful :)
Seems like indent detection behaves radically different if the line starts with punctuation
OK, the good news is that the yak has been shaved and I have a repro. The bad news is that the repro is incredibly touchy, and the issue lies deep in the diff algorithm's "semantic cleanup" code -- i.e., un-grokkable heuristics that are very likely to break other scenarios if I change anything. I'll step through the code tomorrow and see if I get lucky and it's something straightforward where I can predict the side-effects... But probably not.
BTW, this algorithm is not content-sensitive AFAICT but it is extremely sensitive to the structure and length of raw surrounding deltas.
@dhalperi OK, I tried tweaking the algorithm. Basically, what was happening is that short equal deltas that were flanked by longer insertion/deletion deltas were being collapsed into the insertion/deletion. I changed the code to ignore space-only insertions/deletions for the purposes of this computation. I believe it fixes the problem you brought up with hopefully minimal side-effects but I can't be sure. Please keep an eye out for weird regressions!