Issue 347: comment counts are not showing in sidebyside patch history section
Status:  Released
Owner: ----
Closed:  Oct 2012
Reported by stadlera...@gmail.com, Nov 24, 2009
Powered by Gerrit Code Review (v2.0.24)

To reproduce:
See https://android-git.corp.google.com/g/#change,30791
Note that patch set 2 has 24 comments
Now open patch set 3 in side-by-side mode, and click open the Patch 
History.

Expected:
Each patch set will indicate if there were comments

Actual:
Patch set 2 shows zero comments.

Confirmed in both Chrome & FF, on Linux
Nov 25, 2009
#1 sop@google.com
It appears that the patch history table is only showing the number of comments in this 
patch set, and 0 for any other patch set on this file.
Status: Accepted
Feb 23, 2010
#2 sop@google.com
This is actually two different problems.
       
The reason your patch history doesn't show the 24 comments from ps
2 is because the file name changed between the two commits:
       
 ps2: samples/.../MockDemoTest.java              24 comments
  ps3: samples/.../ProviderInjectionDemoTest.java  6 comments
       
Gerrit doesn't grok this ps2-ps3 difference well here.  Its actually
sort of trying to use ps3's file name in order to find the comments
from ps2... which fails.
       
I'm working up a patch to try and improve this for some of the common
cases, but its downright hard without computing all permutations
of patch set against patch set.
       
       
In this particular example you pointed me to, even Git says the
files between ps2 and ps3 are so radically different in content
that they aren't a rename.  Since there was only one file in each
patch set, sure, we could assume they are meant to be a rename and
implicitly force that.  But that won't hold when there are multiple
files in each patch set.  So I'm not really sure how to fix this
in the general case.
 
And before you say, Git's rename detection sucks and explicit 
rename is better... what would you have seen in Google Mondrian  
for this case?  The file was new either way (relative to the base 
branch).  The author dropped one path out of the CL in patch set
2 and added a totally different path into patch set 3.  IIRC even
Mondrian/Perforce wouldn't show this as a rename as there is no
place to get the integration data from.
Status: Started
Labels: Milestone-2.1.2
Feb 23, 2010
#3 sop@google.com
I think change Ibb1d7d5b8b002d0094af8e070b701d2151333de0 does
about as much as we can be expected to do without computing
the full permutations of every patch set against every other
patch set in the same change just to see if there were renames
going on.
Status: Fixed
Labels: -Milestone-2.1.2 FixedIn-2.1.2
Oct 21, 2012
#4 sop@google.com
(No comment was entered for this change.)
Status: Released