cvs commit: ports GIDs UIDs ports/net Makefile ports/net/minidlna Makefile distinfo pkg-descr pkg-plist ports/net/minidlna/files BSDmakefile config.h getifaddr.c minidlna.sh.in patch-2928850 patch-aa patch-ab patch-inline patch-minidlna.c patch-upnphttp ...

[Mikhail T.]

On 17.12.2010 16:15, Doug Barton wrote:
> In no particular order, new rc.d scripts should not be added with the 
> .sh extension. I've attached a patch to the current version, but when 
> the final result is committed it should be minidlna.in.
Thanks for the diff, Doug. Regarding the .sh extension, there is a 
problem... This port has NO_WRKSUBDIR, and its main executable itself is 
called minidlna. If I don't use the .sh extension, the script overwrites 
the binary in WRKDIR (which is the same as WRKSRC in this case).

The explicit PID-file removal was necessary, because I patched the 
software to drop root-privileges after creating the PID-file and opening 
its log (under /var/log). If the PID file is not removed before a 
subsequent start, the program will not overwrite it -- unlike rc.subr, 
it has no smarts to figure out, the PID-file is stale.

Now, in my defense, FWIW, I used the net/mediatomb's rc.d script as a 
working (and accepted) example... Sorry, didn't realize the significance 
of the KEYWORD comment...

And I remind, again, that I asked publicly -- on ports@ -- for help in 
creating the start-up script six months ago. No one has volunteered and 
I figured, it is better to have a port with imperfect start-up script, 
than no port at all... Yours,

    -mi