saintstack commented on a change in pull request #2862:
URL: https://github.com/apache/hbase/pull/2862#discussion_r554232498
##########
File path: dev-support/create-release/release-util.sh
##########
@@ -469,8 +469,18 @@ function generate_api_report {
local timing_token
timing_token="$(start_step)"
# Generate api report.
+ # Filter out some jar types. Filters are tricky. Python regex on
+ # file basename. Exclude the saved-aside original jars... they are
+ # not included in resulting artifact. Also, do not include the
+ # hbase-shaded-testing-util.* jars. This jar is unzip'able on mac
+ # os x as is because has it a META_INF/LICENSE file and then a
+ # META_INF/license directory for the included jar's licenses;
+ # it fails to unjar on mac os x which this tool does making its checks
+ # (Its exclusion should be fine; it is just an aggregate of other jars).
"${project}"/dev-support/checkcompatibility.py --annotation \
org.apache.yetus.audience.InterfaceAudience.Public \
+ -e "original-hbase.*.jar" \
+ -e "hbase-shaded-testing-util.*.jar" \
Review comment:
> I'm thinking the reason to retain the shaded jar is to notice if we
break shading entirely. i.e., a change that adds a new jar to the shaded
package would show in the report as loads of new classes. Perhaps we have other
mechanisms in place already for catching these kinds of errors?
This is a valid concern. Lets fix HBASE-25486 and get the shaded jar back in
the loop.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
