Merge lp:~mterry/ubuntu-system-settings/careful-sockets into lp:ubuntu-system-settings
| Status: | Merged |
|---|---|
| Approved by: | Ken VanDine |
| Approved revision: | 904 |
| Merged at revision: | 904 |
| Proposed branch: | lp:~mterry/ubuntu-system-settings/careful-sockets |
| Merge into: | lp:ubuntu-system-settings |
| Diff against target: |
12 lines (+1/-3) 1 file modified
wizard/ubuntu-system-settings-wizard.conf (+1/-3) |
| To merge this branch: | bzr merge lp:~mterry/ubuntu-system-settings/careful-sockets |
| Related bugs: |
| Reviewer | Review Type | Date Requested | Status |
|---|---|---|---|
| PS Jenkins bot | continuous-integration | Approve | |
| Ken VanDine | Approve | ||
| Michał Sawicz (community) | Approve | ||
|
Review via email:
|
|||
Commit message
Avoid a race condition when cleaning up the wizard's Mir socket. This way we won't delete unity8's Mir socket out from under it.
Description of the change
Avoid a race condition when cleaning up the wizard's Mir socket. This way we won't delete unity8's Mir socket out from under it.
The problem is that the cleanup job stops the wizard, which lets unity8 continue starting (the wizard interrupts unity8 starting). The post-stop job for the wizard looked like:
if ! [ -e "$HOME/$RUN_FILE" ]; then
rm -f "$MIR_SOCKET"
fi
But unity8's job might actually re-define the global upstart value of $MIR_SOCKET by the time this code runs. In fact, $HOME/$RUN_FILE may or may not be created by the time this runs either. Bunch of races.
My solution is to just unconditionally delete the specific Mir socket that the wizard creates, rather than rely on $MIR_SOCKET.
I *think* this will fix bug 1355726. I'm fairly confident that it will fix the part I can reproduce. But I can't reproduce the unity8-
| PS Jenkins bot (ps-jenkins) wrote | # |
Makes sense.