Prev: Devanagari question
Next: Tcl Jobs and Degrees
From: IonutAnghelcovici on 27 Jun 2008 05:01
Hello everybody!

I have a big problem with a TCL script. I will do my best to describe
the behavior:

I have a C program that defines some custom TCL commands. The program
parses an input TCL file.
When it starts the parsing the only custom command is defobj <name>
<alias>. What that command does is to create two new commands: "name"
and an alias to it called "alias".

The startup script does the following:

if { [info commands "custom_command"] == "" } {
source prim_script.tcl
}

The prim_script.tcl is:

set test_var custom_command
....
if { $test_var == "custom_command" } {
{
puts $test_var
defobj custom_command custom_alias
puts $test_var
}
......


When I run the script I get the following output:
custom_command
custom_alias

The crazy thing is that if I change the startup script to:

puts [info commands "custom_command"] == "" } {
source prim_script.tcl
}


Can somebody help me understand why the value of test_var is changed
to cusom_alias ?

Thank you,
Ionut Anghelcovici



From: Donald G Porter on 27 Jun 2008 08:43
IonutAnghelcovici wrote:
> When it starts the parsing the only custom command is defobj <name>
> <alias>. What that command does is to create two new commands: "name"
> and an alias to it called "alias".

> puts $test_var
> defobj custom_command custom_alias
> puts $test_var

> When I run the script I get the following output:
> custom_command
> custom_alias

> Can somebody help me understand why the value of test_var is changed
> to cusom_alias ?

It's hard to be sure, but the results suggest that the [defobj]
command may not be properly respecting the Copy-On-Write rules for
modifying Tcl values.

Is the implementation of [defobj] available for review?

--
| Don Porter Mathematical and Computational Sciences Division |
| donald.porter(a)nist.gov Information Technology Laboratory |
| http://math.nist.gov/~DPorter/ NIST |
|______________________________________________________________________|
From: Alexandre Ferrieux on 27 Jun 2008 09:06
On Jun 27, 2:43 pm, Donald G Porter <d...(a)nist.gov> wrote:
> IonutAnghelcovici wrote:
> > When it starts the parsing the only custom command is defobj <name>
> > <alias>. What that command does is to create two new commands: "name"
> > and an alias to it called "alias".
> > puts $test_var
> > defobj custom_command custom_alias
> > puts $test_var
> > When I run the script I get the following output:
> > custom_command
> > custom_alias
> > Can somebody help me understand why the value of test_var is changed
> > to cusom_alias ?
>
> It's hard to be sure, but the results suggest that the [defobj]
> command may not be properly respecting the Copy-On-Write rules for
> modifyingTclvalues.
>
> Is the implementation of [defobj] available for review?

I've found http://ftp.swarm.org/pub/swarm/src/swarm/swarm-2.2.tar.gz
Apart from the programming langauge being objc, one striking thing is
that it seems to be using the old argc/argc command interface
(Tcl_CreateCommand). To violate COW in this case, the code must be
casting over a "const char *", but I have not yet homed in on the
spot.

-Alex
From: IonutAnghelcovici on 27 Jun 2008 10:35
On Jun 27, 4:06 pm, Alexandre Ferrieux <alexandre.ferri...(a)gmail.com>
wrote:
> On Jun 27, 2:43 pm, Donald G Porter <d...(a)nist.gov> wrote:
>
>
>
> > IonutAnghelcovici wrote:
> > > When it starts the parsing the only custom command is defobj <name>
> > > <alias>. What that command does is to create two new commands: "name"
> > > and an alias to it called "alias".
> > >     puts $test_var
> > >     defobj custom_command custom_alias
> > >     puts $test_var
> > > When I run the script I get the following output:
> > > custom_command
> > > custom_alias
> > > Can somebody help me understand why the value of test_var is changed
> > > to cusom_alias ?
>
> > It's hard to be sure, but the results suggest that the [defobj]
> > command may not be properly respecting the Copy-On-Write rules for
> > modifyingTclvalues.
>
> > Is the implementation of [defobj] available for review?
>
> I've foundhttp://ftp.swarm.org/pub/swarm/src/swarm/swarm-2.2.tar.gz
> Apart from the programming langauge being objc, one striking thing is
> that it seems to be using the old argc/argc command interface
> (Tcl_CreateCommand). To violate COW in this case, the code must be
> casting over a "const char *", but I have not yet homed in on the
> spot.
>
> -Alex


Hello guys and thank you for the quick response!

You are both right. I was using the old command interface and, playing
with some pointers there, I didn't see that I was overwriting the
value of one of the argv[].

After fixing that it works perfectly.

Thank you again,

Ionut
 | 
Pages: 1
Prev: Devanagari question
Next: Tcl Jobs and Degrees