Hi all, I'm seeing some strange behavior with tracing code, which I'd love some help understanding. I've boiled it down to the following simplified example: (define count (lambda (n) (if (equal? n 1) 1 (+ 1 (count (- n 1)))))) ,tr (count 92) When I run the above code (in a fresh Guile instance, via redirecting from a file), the trace runs to completion. However, when I run this version of the code from a file, which runs the function twice... (define count (lambda (n) (if (equal? n 1) 1 (+ 1 (count (- n 1)))))) ,tr (count 10) ; runs to completion ,tr (count 92) ; dies mid-trace ... then the trace dies when it runs the second time, in something that must be some sort of stack overflow, though the error is fairly cryptic: While executing meta-command: In procedure +: Wrong type argument in position 1: #<unspecified> What's happening from a memory perspective where tracing the code the first time affects how much memory is available for the second time? It seems that the first run shouldn't affect the stack limit of the second, but something is going on. Thanks for the help. Dave
Hi Dave,
On Mon, Sep 28, 2020 at 10:06 AM Dave Musicant <dmusicant@carleton.edu>
wrote:
> Hi all,
>
> I'm seeing some strange behavior with tracing code, which I'd love some
> help understanding. I've boiled it down to the following simplified
> example:
>
> (define count
> (lambda (n)
> (if (equal? n 1) 1
> (+ 1 (count (- n 1))))))
>
> ,tr (count 92)
>
>
> When I run the above code (in a fresh Guile instance, via redirecting
> from a file), the trace runs to completion.
>
> However, when I run this version of the code from a file, which runs the
> function twice...
>
> (define count
> (lambda (n)
> (if (equal? n 1) 1
> (+ 1 (count (- n 1))))))
>
> ,tr (count 10) ; runs to completion
> ,tr (count 92) ; dies mid-trace
>
>
> ... then the trace dies when it runs the second time, in something that
> must be some sort of stack overflow, though the error is fairly cryptic:
>
> While executing meta-command:
> In procedure +: Wrong type argument in position 1: #<unspecified>
>
> What's happening from a memory perspective where tracing the code the
> first time affects how much memory is available for the second time? It
> seems that the first run shouldn't affect the stack limit of the second,
> but something is going on.
>
> Thanks for the help.
> Dave
>
I've tried it with Guile 3.0.4 (in macOS) and I didn't get this error. What
version of Guile and OS are you using?
Aleix
I'm using Guile 2.2.7, under Ubuntu 20.04.
--
Dave
On 10/3/20 2:15 AM, Aleix Conchillo Flaqué wrote:
>
> Hi Dave,
>
> On Mon, Sep 28, 2020 at 10:06 AM Dave Musicant <dmusicant@carleton.edu
> <mailto:dmusicant@carleton.edu>> wrote:
>
> Hi all,
>
> I'm seeing some strange behavior with tracing code, which I'd love
> some
> help understanding. I've boiled it down to the following
> simplified example:
>
> (define count
> (lambda (n)
> (if (equal? n 1) 1
> (+ 1 (count (- n 1))))))
>
> ,tr (count 92)
>
>
> When I run the above code (in a fresh Guile instance, via redirecting
> from a file), the trace runs to completion.
>
> However, when I run this version of the code from a file, which
> runs the
> function twice...
>
> (define count
> (lambda (n)
> (if (equal? n 1) 1
> (+ 1 (count (- n 1))))))
>
> ,tr (count 10) ; runs to completion
> ,tr (count 92) ; dies mid-trace
>
>
> ... then the trace dies when it runs the second time, in something
> that
> must be some sort of stack overflow, though the error is fairly
> cryptic:
>
> While executing meta-command:
> In procedure +: Wrong type argument in position 1: #<unspecified>
>
> What's happening from a memory perspective where tracing the code the
> first time affects how much memory is available for the second
> time? It
> seems that the first run shouldn't affect the stack limit of the
> second,
> but something is going on.
>
> Thanks for the help.
> Dave
>
>
> I've tried it with Guile 3.0.4 (in macOS) and I didn't get this error.
> What version of Guile and OS are you using?
>
> Aleix
>
It seems like a bug in 2.2.7 (or even 2.x). I can only suggest reporting it
to bug-guile@gnu.org and guile-devel@gnu.org and update to Guile 3.x.
Aleix
On Sat, Oct 3, 2020 at 1:39 PM Dave Musicant <dmusicant@carleton.edu> wrote:
> I'm using Guile 2.2.7, under Ubuntu 20.04.
>
>
> --
> Dave
>
> On 10/3/20 2:15 AM, Aleix Conchillo Flaqué wrote:
>
>
> Hi Dave,
>
> On Mon, Sep 28, 2020 at 10:06 AM Dave Musicant <dmusicant@carleton.edu>
> wrote:
>
>> Hi all,
>>
>> I'm seeing some strange behavior with tracing code, which I'd love some
>> help understanding. I've boiled it down to the following simplified
>> example:
>>
>> (define count
>> (lambda (n)
>> (if (equal? n 1) 1
>> (+ 1 (count (- n 1))))))
>>
>> ,tr (count 92)
>>
>>
>> When I run the above code (in a fresh Guile instance, via redirecting
>> from a file), the trace runs to completion.
>>
>> However, when I run this version of the code from a file, which runs the
>> function twice...
>>
>> (define count
>> (lambda (n)
>> (if (equal? n 1) 1
>> (+ 1 (count (- n 1))))))
>>
>> ,tr (count 10) ; runs to completion
>> ,tr (count 92) ; dies mid-trace
>>
>>
>> ... then the trace dies when it runs the second time, in something that
>> must be some sort of stack overflow, though the error is fairly cryptic:
>>
>> While executing meta-command:
>> In procedure +: Wrong type argument in position 1: #<unspecified>
>>
>> What's happening from a memory perspective where tracing the code the
>> first time affects how much memory is available for the second time? It
>> seems that the first run shouldn't affect the stack limit of the second,
>> but something is going on.
>>
>> Thanks for the help.
>> Dave
>>
>
> I've tried it with Guile 3.0.4 (in macOS) and I didn't get this error.
> What version of Guile and OS are you using?
>
> Aleix
>
>
>
>
>