we make heavy use of svn externals, so the svn information for a project with 2 externals, for which 3 issues were affected would look like this on the build log page:

Revisions

• xxx/yyy : 313
• xxx/zzz : 848

Changes

1. #JIRA-37
#JIRA-38
#JIRA-40
2. #JIRA-37
#JIRA-38
#JIRA-40

1 and 2 are exactly the same, since yyy and zzz share a common base.
As a result, the comment that gets comitted by the JIRA plugin contains 2 times the exact same text, which is very, very disturbing and unnecessary. It also contains 2 times the exact same filelist.. Example for the above case:

Integrated in xxx #50 (See [...])
#JIRA-37
#JIRA-38
#JIRA-40
#JIRA-37
#JIRA-38
#JIRA-40

user :
Files :
*/xxx/yyy/file1
*/xxx/yyy/file2
*/xxx/zzz/file1

user :
Files :
*/xxx/yyy/file1
*/xxx/yyy/file2
*/xxx/zzz/file1

Imagine this for 20 externals, 30 changed files and a comment spreading over 10 lines. Yes, that's over 800 lines, where 40 lines would be sufficient.

It should not be too hard to parse the svn info and extract only those comments that are different.