#13 fileevent readable failure

[Anonymous]

revision: Memchan 2.2.1

Use of fileevent readable with fifo fails

reference case with memchan
-------------------------------------
package require Memchan

proc isReadable { f } {
if {![eof $f]} {
puts [read $f]
}
}

set fid [memchan]
fileevent $fid readable [list isReadable $fid]

after 5 {puts $::fid "successful" ; seek $::fid 0}
after 500 {set ::DONE 1}

vwait ::DONE
close $fid
----------------------------------------

failure case with fifo
------------------------------------
package require Memchan

proc isReadable { f } {
if {![eof $f]} {
puts [read $f]
}
}

set fid [fifo]
fileevent $fid readable [list isReadable $fid]

after 5 {puts $::fid "fail"}
after 500 {set ::DONE 1}

vwait ::DONE
close $fid

[Pat Thoyts]

Logged In: YES
user_id=202636

This looks like the same bug I've been looking into with the
rechan extension from tclkit. In my case I've been using
fcopy and finding that fcopying from a channel with no data
causes a hang. A good example is

package require Memchan
set done doing
set src [fifo]
#puts $src \x00 ;# any data in the channel fixes the problem.
set dst [zero]
proc Complete {args} {set ::done complete}
proc Timeout {} {set ::done timeout}
set aid [after 2000 Timeout]
fcopy $src $dst -command Complete
vwait ::done
after cancel $aid
close $src; close $dst
puts "Done: $done"

It looks like a general failure in all the memchan channels.
Are we missing a flag somewhere? The only setting thats
apparent to me is that a short read > 0 sets a blocking flag
in the upper channel layer that seems to cause the next call
to the WatchProc to have mask be non-zero. This is
sufficient to set the timer and ensure we get another
fileevent raised with the eof set.

[Nobody/Anonymous]

Sourceforge.. Reposted it :)