TCL HTTP: can't read "state(after)"

[De Kus]

I wonder if I do something wrong. Full error:

Code: Select all

[-RO^Bot-] [12:58:59] #De_Kus# set errorInfo
[-RO^Bot-] Currently: can't read "state(after)": no such element in array
[-RO^Bot-] Currently:     while executing
[-RO^Bot-] Currently: "after cancel $state(after)"

This occurs with most (all?) asyncron HTTP requests I use in eggdrop. The general problem is obvious: after using ::http::cleanup $token, $state(after) has already been removed before its called by HTTP. But what else should I do? Do you believe using utimer 1 with that command would be a solution or rather a workaround? Is that an error of the HTTP package and should be reported to tcl.tk?

If greped this part of the script:

Code: Select all

[-RO^Bot-] [14:02:58] tcl: evaluate (.tcl): exec grep after /usr/local/lib/tcl8.4/http2.5/http.tcl
[-RO^Bot-] Tcl:     catch {after cancel $state(after)}

no nice way imho, they could first check if exists state(after) to prevent occurance of unneccessary TCL errors.
This occurs at least till HTTP 2.5.2 from TCL 8.4.12.

Edit: not even after 60000 for http::cleanup or a set state(after) "" after the cleanup helps...

[Alchera]

I gather nothing further has been done to fix this (not that I'm surprised)?

De Kus: Did you lodge a bug report?

[rosc2112]

I've only seen that error when calling ::http::cleanup directly after an error test, like in:

Code: Select all

catch {set page [::http::geturl $movieurl -timeout $::movietimeout]} error
	if {[string match -nocase "*couldn't open socket*" $error]} {
		puthelp "PRIVMSG $chan :Error: couldn't connect to filmspot.com..Try again later."
      ::http::cleanup $page
		return
	}
	if { [::http::status $page] == "timeout" } {
		puthelp "PRIVMSG $chan :Error: Connection to filmspot.com timed out..Try again later."
      ::http::cleanup $page
		return
	}
	set html [::http::data $page]

That's how I used to do my web procs. Now I just leave off the http::cleanup call.

I posted about the same error messages before, and was informed that http::cleanup is on a timer anyway, so calling it directly was unneeded. Haven't seen that error in the scripts I fixed that don't call http::cleanup (except at the end, after getting data.)

IOW, my newer scripts take the form:

Code: Select all

catch {set page [::http::geturl $movieurl -timeout $::movietimeout]} error
	if {[string match -nocase "*couldn't open socket*" $error]} {
		puthelp "PRIVMSG $chan :Error: couldn't connect to filmspot.com..Try again later."
		return
	}
	if { [::http::status $page] == "timeout" } {
		puthelp "PRIVMSG $chan :Error: Connection to filmspot.com timed out..Try again later."
		return
	}
	set html [::http::data $page]
	::http::cleanup $page

[Zircon]

Code: Select all

<Anti-Flood> [03:06] #Owner# set errorInfo
<Anti-Flood> Currently: can't read "state(after)": no such element in array
<Anti-Flood> Currently:     while executing
<Anti-Flood> Currently: "after cancel $state(after)"

i always was trying to identify the script causing this error. So finally i understand it s related to the HTTP. Indeed, i use many scripts that deals with HTTP.

[speechles]

no nice way imho, they could first check if exists state(after) to prevent occurance of unneccessary TCL errors.

Testing this theory it seems to work...

lines 165,166 change to:

Code: Select all

    if {[info exists state(sock)]} { catch {close $state(sock)} }
    if {[info exists state(after)]} { catch {after cancel $state(after)} }

lines 197,198 change to:

Code: Select all

    if {[info exists state(sock)]} { catch {fileevent $state(sock) readable {}} }
    if {[info exists state(sock)]} { catch {fileevent $state(sock) writable {}} }

Using HTTP 2.4.2 and changing the lines above it seems to solve the problem. Running various http queries and a quick .set errrorInfo afterwards yields the following...

<speechles> !g google
<sp33chy> 1,460,000,000 Results | Google @ http://www.google.com/ | Google Maps @ http://maps.google.com/ | Google Video @ http://video.google.com/ | Google Groups @ http://groups.google.com/
*meanwhile on the partyline directly after*
<speechles> .set errorInfo
<sp33chy> Currently: couldn't create error file for command: no such file or directory
<sp33chy> Currently: while executing
<sp33chy> Currently: "exec chmod 700 $nb_dir"

<speechles> !w .jasd asd
<sp33chy> Socket Error accessing 'jasd.wikipedia.org' .. Does it exist?
*meanwhile on the partyline directly after*
<speechles> .set errorInfo
<sp33chy> Currently: couldn't open socket: host is unreachable
<sp33chy> Currently: while executing
<sp33chy> Currently: "::http::geturl $query -timeout [expr 1000 * 5]"

Notice the top error message which is usually a state(after) error is instead an error message reporting no error found to report, so this does work as a fix. If someone can shed some light if this is indeed an http package bug it would be wonderful.

[Alchera]

I use only one script that grabs data from the web (and causes this annoyance), the OV2 version of B0unTy's iMDB.tcl.

It uses:

Code: Select all

::http::Finish $page