command pipelines, blocking and stderr

Hello,

I've written a bit of code that is quite useful for launching commands
from Tk without locking up the application (at the end of the message),
but there's something I don't quite get. The code to close things down
is like so:

if { [eof $fl] } {
fconfigure $fl -blocking 1
catch { close $fl } err

which is saying that if there is nothing more to read from the
processes's stdout, we need to close things down and report back to
whoever called us. The problem I was encountering was that without
setting the channel back to blocking before closing it, it didn't report
in 'err' the command's stderr. I sort of see the logic behind this, but
I wanted to enquire about it to see if others feel that this is the
right way to do things...

Opinions?

Thanks,
--
David N. Welton
- http://www.dedasys.com/davidw/

Linux, Open Source Consulting
- http://www.dedasys.com/

------------------

namespace eval runcommand {}

proc ::runcommand::runcommand {cmd} {
variable standarderror
variable standardout

#puts "running $cmd"

set standardout ""
set standarderror ""

set fl [open "| $cmd" r]
fconfigure $fl -blocking 0
fconfigure $fl -buffering none


::runcommand::runcommandloop $fl
vwait ::runcommand::finished

#puts "stdout: $standardout"
#puts "stderr: $standarderror"
return "Output:\n$standardout\nErrors:\n$standarderror"
}

proc ::runcommand::runcommandloop {fl} {
variable standarderror
variable standardout

append standardout [read $fl]
if { [eof $fl] } {
fconfigure $fl -blocking 1
catch { close $fl } err
append standarderror $err
after 1 [list set ::runcommand::finished 1]
return
}
after 100 [list ::runcommand::runcommandloop $fl]
}

set out [runcommand::runcommand "make aaaa"]

puts $out
.