Re: [bug-gawk] gawk 4.1 windows: application crash while dumping SYMTAB array

[Denis Shirokov]

Hi again. Glad to help you. Actually I don't have a skills to apply
this patch. I have another issue with the `exit' statement. GAWK is
also crashing but the problem is in that I cannot simplify the source.
My gawk scripts is not smallest ones. Do You interested in gawk script
 about 200KB length that is 100% reproducing the problem?

regards
Denis Shirokov

2013/5/27, Aharon Robbins <address@hidden>:
> Hello.  Thanks for the bug report.
>
>> Date: Mon, 27 May 2013 19:06:53 +0300
>> From: Denis Shirokov <address@hidden>
>> To: bug-gawk <address@hidden>
>> Subject: [bug-gawk] gawk 4.1 windows: application crash while dumping
>> SYMTAB array
>>
>> Hi Aharon, Eli,
>>
>> Here is my gawk crash report that i was saying few days ago. Sorry
>> this is not simpliest example, may be later i will send you more
>> simply example.
>>
>> Reproducible in Win7x64SP1 and Win2008R2SP1. Other systems i was not
>> tested.
>>
>> The attached file _main.gwk contains program that doing dump of the
>> array FUNCTAB - and it is successed, and then while dumping SYMTAB -
>> it is crashing.
>>
>> Also i attach archive report.piz with the gawk that i was used. This
>> is normal ZIP-archive that i was renamed due to mail-engine is not
>> allowing to send through email executable files, and also inside the
>> archives.
>>
>> Hope this will help You.
>>
>> best Regards
>
> This script crashes on GNU/Linux also.  It is likely the same crash that
> you're seeing on Windows although Eli will have to confirm that.
>
> The following patch fixes it. I believe it is a safe fix.
>
> Arnold
> -----------------------------------
> diff --git a/array.c b/array.c
> index 5dac7a4..92a1cb8 100644
> --- a/array.c
> +++ b/array.c
> @@ -113,7 +113,16 @@ null_array(NODE *symbol)
>       symbol->table_size = symbol->array_size = 0;
>       symbol->array_capacity = 0;
>       symbol->flags = 0;
> -     assert(symbol->xarray == NULL);
> +     /*
> +      * 5/2013: This used to be
> +      *
> +      *      assert(symbol->xarray == NULL);
> +      *
> +      * But that seems to cause problems for no good reason
> +      * that I can see. I believe it to be an artifact of the
> +      * union getting in the way.
> +      */
> +     symbol->xarray = NULL;
>       /* vname, parent_array not (re)initialized */
>  }
>
>