Forum OpenACS Development: callbacks and plain parameters

is there a known problem with callbacks in 5.2 where the formal parameter names are not getting into the appropriate stack frame?

specifically...

The definition:

ad_proc \
    -callback bug_tracker::bug__add_new_log_entry \
    {log_id user_id peeraddr context_id} \
    { allows subscribing code to add db items associated with new log entry } \
    -

An Implementation:

ad_proc \
    -callback bug_tracker::bug__add_new_log_entry \
    -impl action_elapsed_time \
    {log_id user_id peeraddr context_id} \
    { add an ipt and attach it to the log entry of the bug } \
    {
        ...
        ns_log notice "log_id $log_id user_id $user_id peeraddr $peeraddr context_id $context_id"
        ...
    }

The Hook or Call:

set bug__addnewentry_list \
    [callback bug_tracker::bug__add_new_log_entry \
        $bt_rev_id \
        $user_id \
        $peeraddr \
        $package_id] 

I'm getting the error "can't read "log_id": no such variable". Why? Doesn't this suggest a bug with parameters?

I've committed a fix on head which should solve this problem. It only shows up when a contract only has positional arguments. For a plain ad_proc with only positional arguments an arg parser is not created because the standard tcl proc command will happily parse positional arguments for us. For a callback however the implementation relies on the arg parser of the contract so we need to create it anyway.

or, as discussed, to have ad_proc build an arg parser in the case of a callback with only positional parameters.

Dave,

To clarify my reasoning for exposing the patch, Lee Denison had committed this to HEAD earlier (with a few other changes, like ... eq ... replacing [string equal ... ...]). What I did was take his patch and remove everything but the one thing that did the job.

If I understand correctly, feature adds to 5.2 are frozen. My intent was not to change 5.2 as shipped, but rather to provide folks running 5.2 with a working callback functionality if they personally wanted it; if so, they would apply the patch and restart their server.

To clarify what the patch does, it can be seen by looking at the patch itself: the else clause of that structure causes an argument parser to be built. The change adds an additional condition which must be satisfied to -not- build an argument parser: the ad_proc must not be defining a callback.