H
Henry Townsend
Please forgive me for asking here, I have done some searches and come up
empty but I still think the module I need may have been written and want
to know for sure before I go trying to roll my own.
I'm thinking of writing a module which would monitor the output of any
script which "use"d it looking for errors. E.g. something like this
use Output::Monitor STDOUT => qr/re1/, STDERR => qr/re2/;
This would - perhaps by using tied filehandles - watch all data flowing
to stdout and stderr, apply the specified REs to each line, and convert
the exit status to nonzero if any matches occur. The data would still be
delivered to the same place, just checked along the way.
The idea is to replace a common technique where I work, which is to
write build scripts which divert their own output to a log internally,
then read that logfile to discover any unflagged errors. This makes it
hard for external tools to redirect output as they choose, requires
intimate knowledge of the script to know where it's writing its logfile
to, etc. What I'd hope to achieve with the above is transparency, i.e.
by simply including the module within a script it would be converted
from unchecked to checked, without the script having to change anything
else while the same data as before flows to stdout and stderr.
Before anyone says "just check exit status" - we do but we're stuck with
certain underlying build tools which have unreliable status so we must
parse output as a belt-and-suspenders plan.
Does anyone know of such a module or something close to use as a
starting point?
empty but I still think the module I need may have been written and want
to know for sure before I go trying to roll my own.
I'm thinking of writing a module which would monitor the output of any
script which "use"d it looking for errors. E.g. something like this
use Output::Monitor STDOUT => qr/re1/, STDERR => qr/re2/;
This would - perhaps by using tied filehandles - watch all data flowing
to stdout and stderr, apply the specified REs to each line, and convert
the exit status to nonzero if any matches occur. The data would still be
delivered to the same place, just checked along the way.
The idea is to replace a common technique where I work, which is to
write build scripts which divert their own output to a log internally,
then read that logfile to discover any unflagged errors. This makes it
hard for external tools to redirect output as they choose, requires
intimate knowledge of the script to know where it's writing its logfile
to, etc. What I'd hope to achieve with the above is transparency, i.e.
by simply including the module within a script it would be converted
from unchecked to checked, without the script having to change anything
else while the same data as before flows to stdout and stderr.
Before anyone says "just check exit status" - we do but we're stuck with
certain underlying build tools which have unreliable status so we must
parse output as a belt-and-suspenders plan.
Does anyone know of such a module or something close to use as a
starting point?