Tcl Source Code

% proc stack {} {
    puts LEVEL:[set lev [info frame]]
    for {set i 1} {$i < $lev} {incr i} {
puts [info frame $i]
    }
}
% stack
LEVEL:2
type eval line 1 cmd stack level 1
% proc a {} stack
% a
LEVEL:3
type eval line 1 cmd a level 2
type proc line 1 cmd stack proc ::a level 1
% coroutine aa a
LEVEL:5
bad level "1"

allow_comments - 1

That did it; committed.

Assuming the alternative is true the fix would be to remove the +1 in line 1103 of file tclCmdIL.c, rev 1.174.

An alternative is that the calculation of levels in the coroutine is off by one, i.e. that info frame should show only the coro stack up to what we see in my output as level 2.

IP - Comment Removed: 204.244.102.2

I confirm, using my own frame dumper (which starts from the current from down to 1, so I am able to see some output before it croaks, my code is attached here). Output is

andreask@gila:~/workbench/Coro280> ./install/tcl86/bin/tclsh8.6
% source infoframedebug.tcl
% debugframes
FRAME   2 {type source line 6 file infoframedebug.tcl cmd {info frame $level} proc ::debugframes level 0}
FRAME   1 {type eval line 1 cmd debugframes level 1}

% proc a {} debugframes
% a
FRAME   3 {type source line 6 file infoframedebug.tcl cmd {info frame $level} proc ::debugframes level 0}
FRAME   2 {type proc line 1 cmd debugframes proc ::a level 1}
FRAME   1 {type eval line 1 cmd a level 2}

% coroutine aa a
FRAME   5 {type source line 6 file infoframedebug.tcl cmd {info frame $level} proc ::debugframes level 0}
FRAME   4 {type proc line 1 cmd debugframes proc ::a level 1}
FRAME   3 {type eval line 1 cmd a}
FRAME   2 {type eval line 1 cmd {coroutine aa a} level 2}
bad level "1"
% exit

It seems the trouble is where it tries to access the stack frames above the coroutine invokation.

The bad level is either thrown for out of range, or, for in-range numbers, if the loop in lines 1142-1148 (tclCmIL.c, rev 1.174) is not finding a frame. That would indicate that the framePtr->nextPtr chain is broken at the point of main to coro transition.

This is likely managed in the code switching between coros, right ?

I confirm, using my own frame dumper (which starts from the current from down to 1, so I am able to see some output before it croaks, my code is attached here). Output is

andreask@gila:~/workbench/Coro280> ./install/tcl86/bin/tclsh8.6
% source infoframedebug.tcl
% debugframes
FRAME   2 {type source line 6 file infoframedebug.tcl cmd {info frame $level} proc ::debugframes level 0}
FRAME   1 {type eval line 1 cmd debugframes level 1}

% proc a {} debugframes
% a
FRAME   3 {type source line 6 file infoframedebug.tcl cmd {info frame $level} proc ::debugframes level 0}
FRAME   2 {type proc line 1 cmd debugframes proc ::a level 1}
FRAME   1 {type eval line 1 cmd a level 2}

% coroutine aa a
FRAME   5 {type source line 6 file infoframedebug.tcl cmd {info frame $level} proc ::debugframes level 0}
FRAME   4 {type proc line 1 cmd debugframes proc ::a level 1}
FRAME   3 {type eval line 1 cmd a}
FRAME   2 {type eval line 1 cmd {coroutine aa a} level 2}
bad level "1"
% exit

It seems the trouble is where it tries to access the stack frames above the coroutine invokation.

The bad level is either thrown for out of range, or, for in-range numbers, if the loop in lines 1142-1148 (tclCmIL.c, rev 1.174) is not finding a frame. That would indicate that the framePtr->nextPtr chain is broken at the point of main to coro transition.

This is likely managed in the code switching between coros, right ?

File Added - 354261: infoframedebug.tcl