pipe-tools-0.45 compile

pipe-tools-0.45 compile

[Taj Khattra]

hi,

pipe-tools-0.45 doesn't compile with the latest skalibs.
i think nptrigger_make() needs to be modified to use
sauniquename() (UNIQUENAME is no longer defined in skalibs).

also, would it make any sense to have killchildexit() in
maximumtime.c use wait_pid() instead of wait_nohang() to
prevent a potential zombie (e.g. in case the child is ignoring
SIGTERM) ?

thanks

-taj

Re: pipe-tools-0.45 compile

[Laurent Bercot]

> pipe-tools-0.45 doesn't compile with the latest skalibs.
> i think nptrigger_make() needs to be modified to use
> sauniquename() (UNIQUENAME is no longer defined in skalibs).

 pipe-tools essentially needs to be rewritten from scratch. The
0.45 libnptrigger is very ugly and uselessly resource-consuming.
pipe-tools 0.45 works (with previous versions of skalibs), but
I don't advocate its use.
 A brand new shiny version is slowly - but surely - on the way.
Don't expect it tomorrow, though.


> also, would it make any sense to have killchildexit() in
> maximumtime.c use wait_pid() instead of wait_nohang() to
> prevent a potential zombie (e.g. in case the child is ignoring
> SIGTERM) ?

 No: maximumtime is supposed to exit after the given number of
seconds. The wait_nohang() call is basically useless: I left it
there because it's cleaner to reap zombies yourself, if you have any,
than to leave that work to process 1. But it's not really needed.
Really waiting for the child with wait_pid(), OTOH, would break the
guarantee that maximumtime exits after n seconds. If the child doesn't
die, too bad - it's the child's responsibility to handle SIGTERM properly.

-- 
 Ska