| 1 |
I've noticed that some init scripts call start-stop-daemon when the |
| 2 |
"daemon" is really a shell script. This needs to be changed so that |
| 3 |
either |
| 4 |
|
| 5 |
1) The call to start-stop-daemon is removed in the init script. |
| 6 |
This is logical as shell scripts are not daemons |
| 7 |
|
| 8 |
2) Embed the functionality of the shell script in the init script so |
| 9 |
that the start-stop-daemon command really does launch the daemon. |
| 10 |
|
| 11 |
This change is required as future versions of baselayout will wrap the |
| 12 |
start-stop-daemon command to test that it has actually started and is |
| 13 |
still running. If it's not running then return failure instead of |
| 14 |
returning success as it does at the moment. Obviously this will fail if |
| 15 |
we're trying to launch a shell script! |
| 16 |
|
| 17 |
This solves many bugs where daemons launch fine and return success to |
| 18 |
start-stop-daemon but then die due to an invalid configuration file - |
| 19 |
named/bind is a good example of this. |
| 20 |
|
| 21 |
The only ebuild that I know of that requires this change is |
| 22 |
net-mail/courier-imap |
| 23 |
|
| 24 |
There may be others in the tree - only the ebuild maintainers will know. |
| 25 |
If you're an ebuild maintainer and install an init script, please check |
| 26 |
this otherwise when a baselayout-1.12.0 hits things will break. |
| 27 |
|
| 28 |
Another thing to stop doing is blindly rm -f pidfiles without testing to |
| 29 |
see if they exist or not as we now clean up pidfiles for you - if we |
| 30 |
can. This won't cause any problems other than an error message, so it's |
| 31 |
not a major issue, but I thought I'd let people know. |
| 32 |
|
| 33 |
Thanks |
| 34 |
|
| 35 |
-- |
| 36 |
Roy Marples <uberlord@g.o> |
| 37 |
Gentoo Linux Developer |
Archives