Wrote a little too soon. This patch seems to correctly report the caller
whereas before it didn't. However I do think one still wants locations
inside the trap.

trap DEBUG is a little bit of a misnomer. trap TRACE would be more accurate
and it can be used for example in profiling code. Personally I would like to
be able to debug a debugger, but I realize that's perhaps a bit esoteric
(although in Ruby one can have nested hooks).

But suppose one wrote a profiler, or just some sort of signal handler. Don't
you want to be able to debug via PS4 the same as any other code?

On Thu, Sep 18, 2008 at 11:05 AM, Rocky Bernstein <rocky.bernstein@gmail.com
> wrote:

> Thanks for fixing.
>
> On Thu, Sep 18, 2008 at 8:13 AM, Peter Stephenson <pws@csr.com> wrote:
>
>> On Wed, 17 Sep 2008 12:39:36 -0400
>> "Rocky Bernstein" <rocky.bernstein@gmail.com> wrote:
>> > On Tue, Sep 16, 2008 at 10:50 AM, Peter Stephenson <pws@csr.com> wrote:
>> >
>> > > Now we have logic for finding the source file and corresponding line
>> > > number of executed code, this adds the prompt escapes %x and %I which
>> > > are like %N and %i but for the file where the code was defined.
>> >
>> > Looks like this has problems when used inside trap DEBUG.
>>
>> What's happening here is that the current shell logic doesn't update line
>> numbers when inside a trap, even inside nested functions.
>
>
> Ok. Thanks for the explanation. I don't see how this gives a line number
> that is larger than the number of lines in the entire program, but if the
> problem is fixed this is moot.
>
>
>> The idea is that
>> you're interested in the line number of what triggered the trap; this is
>> particularly true of a DEBUG trap where you might want to examine LINENO
>> to
>> see what's going on in the code you're debugging, and you're much less
>> likely to be interested in the line number inside the trap itself.
>
>
> If one wants the line number that triggered the trap, there are two simple
> ways to get this:
> 1. Save the line number in the call
>    trap 'int_handler $LINENO'  INT
> 2. Use funcfiletrace
>
>
>>
>> So the obvious fix is to make the prompt code I added take account of
>> this.
>
>
> Yes, I think this the best thing.
>
>
>>
>>
>> Another possibility would be to make traps respect the EVALLINENO option,
>> so that you would get the local line numbers.  However, I think that's
>> less
>> useful.
>>
>> It's possible the same problem affects some of the other new features.  In
>> particular I think funcstack line number info for eval commands inside
>> traps is going to get thoroughly confused.  I think the simplest fix would
>> be not to add funcstack info for eval in this case; this is sort of
>> plausible since it's like the case where eval isn't keep track of its own
>> line numbers, where we don't add funcstack info, and we don't add
>> funcstack
>> info for a trap which is like an eval without its own line number.
>> Otherwise it's not currently obvious to me what the fix would be.
>>
>> It might be a good idea to add special variables ZSH_SOURCE_LINENO and
>> ZSH_SOURCE_FILE as more readable alternatives to ${(%):-%I} and
>> ${(%):-%x}.
>>
>> Index: Src/prompt.c
>> ===================================================================
>> RCS file: /cvsroot/zsh/zsh/Src/prompt.c,v
>> retrieving revision 1.54
>> diff -u -r1.54 prompt.c
>> --- Src/prompt.c        16 Sep 2008 15:07:16 -0000      1.54
>> +++ Src/prompt.c        18 Sep 2008 11:55:04 -0000
>> @@ -726,7 +726,8 @@
>>                     stradd(Rstring);
>>                break;
>>            case 'I':
>> -               if (funcstack && funcstack->tp != FS_SOURCE) {
>> +               if (funcstack && funcstack->tp != FS_SOURCE &&
>> +                   (!intrap || trapisfunc)) {
>>                    /*
>>                      * We're in a function or an eval with
>>                     * EVALLINENO.  Calculate the line number in
>> @@ -749,7 +750,8 @@
>>                 bp += strlen(bp);
>>                break;
>>            case 'x':
>> -               if (funcstack && funcstack->tp != FS_SOURCE)
>> +               if (funcstack && funcstack->tp != FS_SOURCE &&
>> +                   (!intrap || trapisfunc))
>>                     promptpath(funcstack->filename ? funcstack->filename :
>> "",
>>                               arg, 0);
>>                else
>> --
>> Peter Stephenson <pws@csr.com>                  Software Engineer
>> CSR PLC, Churchill House, Cambridge Business Park, Cowley Road
>> Cambridge, CB4 0WZ, UK                          Tel: +44 (0)1223 692070
>>
>
>