[MDEV-18970] uninited var can be read in gtid_delete_pending() Created: 2019-03-19 Updated: 2019-05-16 Resolved: 2019-05-16 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Replication |
| Affects Version/s: | 10.4.1 |
| Fix Version/s: | 10.4.5 |
| Type: | Bug | Priority: | Major |
| Reporter: | Andrei Elkin | Assignee: | Sujatha Sivakumar (Inactive) |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Description |
|
marko reports
the warning: The following patch should suffice:
as there's no urgency to call gtid_check_rpl_slave_state_table() earlier than the var-in-question gets assigned. |
| Comments |
| Comment by Sujatha Sivakumar (Inactive) [ 2019-03-29 ] |
|
Hello Andrei, Please review the fix for http://buildbot.askmonty.org/buildbot/grid?category=main&branch=bb-10.4-sujatha To reproduce I followed these instructions: sujatha@sujatha:~/bug_repo/ =OFF -DWITH_EMBEDDED_SERVER=OFF -DWITH_UNIT_TESTS=OFF -DCMAKE_BUILD_TYPE=Debug -DPLUGIN_TOKUDB=NO -DPLUGIN_MROONGA=NO -DPLUGIN_OQGRAPH=NO -DPLUGIN_ROCKSDB=NO -DPLUGIN_CONNECT=NO -DWITH_SSL=bundled .. Please let me know if you need additional details. Thanks. |
| Comment by Sujatha Sivakumar (Inactive) [ 2019-03-29 ] |
|
Fixed the patch specific issue in my earlier version Andrei. Tested the changes locally. |
| Comment by Andrei Elkin [ 2019-05-15 ] |
|
Sujatha, salute. The patch looks good. I would suggest to assert instead of comment in the last hunk. Thanks for this! Andrei |