Is this a bug in overlay property list handling?

Is this a bug in overlay property list handling?

[Lennart Borgman (gmail)]

I am not sure if I have hit a bug or if I am doing something wrong. I do 
something like this in my code:

   (setq syntax-ppss-last-min
         (copy-tree
          (cons (1- last-pos)
                (if is-main-mode-chunk
                    (mumamo-with-major-mode-fontification main-major
                      `(parse-partial-sexp (point-min) last-pos))
                  (parse-partial-sexp 1 1)))
          ))
   (overlay-put chunk-at-pos 'syntax-ppss-last-min syntax-ppss-last-min)

If I test this immediately after with `overlay-get' the value looks ok.

However, later when I read these values with `overlay-get' it seems like 
the property list has got corrupted. The car of the value stored in 
property 'syntax-ppss-last-min is nil.

I use copy-tree to try to avoid surprices like this one. Maybe I should 
do something different? Maybe this is a bug in overlay property list 
handling?

Re: Is this a bug in overlay property list handling?

[Stefan Monnier]

> I am not sure if I have hit a bug or if I am doing something wrong. I do
> something like this in my code:

>   (setq syntax-ppss-last-min
>         (copy-tree
>          (cons (1- last-pos)
>                (if is-main-mode-chunk
>                    (mumamo-with-major-mode-fontification main-major
>                      `(parse-partial-sexp (point-min) last-pos))
>                  (parse-partial-sexp 1 1)))
>          ))
>   (overlay-put chunk-at-pos 'syntax-ppss-last-min syntax-ppss-last-min)

> If I test this immediately after with `overlay-get' the value looks ok.

> However, later when I read these values with `overlay-get' it seems like the
> property list has got corrupted. The car of the value stored in property
> syntax-ppss-last-min is nil.

> I use copy-tree to try to avoid surprices like this one.

The `cons' already makes sure you have the only reference to the object
and parse-partial-sexp also returns a fresh value, so the `copy-tree' is
not needed.

> Maybe I should do something different? Maybe this is a bug in overlay
> property list handling?

I'd be very surprised if it's a bug in overlay property handling.
My guess is that it's a bug elsewhere in your code.


        Stefan


PS: BTW, please avoid using `1' instead of (point-min), or else add
a comment explaining why (point-min) is not right.  Also the
`(parse-partial-sexp (point-min) last-pos) looks odd: you probably
forgot a comma before `last-pos'.
Maybe mumamo-with-major-mode-fontification should be a macro rather than
a function.

Re: Is this a bug in overlay property list handling?

[Lennart Borgman (gmail)]

Stefan Monnier wrote:
>> I am not sure if I have hit a bug or if I am doing something wrong. I do
>> something like this in my code:
> 
>>   (setq syntax-ppss-last-min
>>         (copy-tree
>>          (cons (1- last-pos)
>>                (if is-main-mode-chunk
>>                    (mumamo-with-major-mode-fontification main-major
>>                      `(parse-partial-sexp (point-min) last-pos))
>>                  (parse-partial-sexp 1 1)))
>>          ))
>>   (overlay-put chunk-at-pos 'syntax-ppss-last-min syntax-ppss-last-min)
> 
>> If I test this immediately after with `overlay-get' the value looks ok.
> 
>> However, later when I read these values with `overlay-get' it seems like the
>> property list has got corrupted. The car of the value stored in property
>> syntax-ppss-last-min is nil.
> 
>> I use copy-tree to try to avoid surprices like this one.
> 
> The `cons' already makes sure you have the only reference to the object
> and parse-partial-sexp also returns a fresh value, so the `copy-tree' is
> not needed.

Thanks for the explanation.

>> Maybe I should do something different? Maybe this is a bug in overlay
>> property list handling?
> 
> I'd be very surprised if it's a bug in overlay property handling.
> My guess is that it's a bug elsewhere in your code.

Ok, this should make it easier to find this bug.


>         Stefan
> 
> 
> PS: BTW, please avoid using `1' instead of (point-min), or else add
> a comment explaining why (point-min) is not right.

I will add a comment.

> Also the
> `(parse-partial-sexp (point-min) last-pos) looks odd: you probably
> forgot a comma before `last-pos'.

Yes, I did. Thanks.

> Maybe mumamo-with-major-mode-fontification should be a macro rather than
> a function.

I am not sure now. I will have a look again.

Re: Is this a bug in overlay property list handling?

[Lennart Borgman (gmail)]

>> Also the
>> `(parse-partial-sexp (point-min) last-pos) looks odd: you probably
>> forgot a comma before `last-pos'.
> 
> Yes, I did. Thanks.


This was the bug. It works as I expected after adding the comma.