Hi Joro,
A few comments on the patch.
On 19.07.10 19:12, Georgi.Kodinov@stripped wrote:
> #At file:///C:/mysql/work/B52515-5.1-bugteam/ based on
> revid:jon.hauglid@stripped
>
> 3492 Georgi.Kodinov@stripped 2010-07-19
> Bug #52515: mysql sessions are not terminated properly
>
> The mysql command line client was not reacting at all on
> closing its controlling terminal.
It was. The default action for SIGHUP on Unix-like systems is to
terminate the process.
> Now it intercepts the SIGHUP on unix-alikes (and the
> corresponding windows way of notification) and issues
> a KILL command before closing similarly to how Ctrl-C
> is handled.
Not exactly. Since you are reusing the same signal handler that is used
for SIGINT, the actual behavior on Unix-like systems now is:
1. If no query is running, just quit on SIGHUP.
2. Otherwise, issue a KILL QUERY when SIGHUP is delivered and then
continue execution.
(I don't consider cases of repeated SIGHUP deliveries)
I don't think that's what you meant it to be.
> Added a command line option to turn this new behaviour
> off if needed.
>
With --sighup-ignore we actually get a new behavior too. Instead of
terminating the client process (and leaving a query running, if any),
mysql started with --sighup-ignore now ignores the signal, i.e. both the
query and the client process are left running.
I vote on removing this option from the patch. I doubt there is anyone
exploiting the old behavior (terminate process, leave query), and if
anyone wants the new --sighup-ignore behavior there are standard OS
facilities (nohup on Unix, scheduled tasks on Windows).
Best regards,
Alexey.
