Re: [Libunwind-devel] [PATCHv3 1/4] Add core dump unwinding support to libunwind

[Arun Sharma]

On Mon, Feb 27, 2012 at 7:54 AM, Denys Vlasenko <address@hidden> wrote:
> On 02/27/2012 04:53 PM, Denys Vlasenko wrote:
>>
>> Patch 1:
>> Introduce struct elf_dyn_info: a common part of struct UPT_info
>> and struct UCD_info (to be introduced later).
>> Make _UPTi_find_unwind_table function operate only on this part
>> of struct UPT_info.

This patch doesn't apply  cleanly to HEAD (commit 3d08506).
Specifically, I have problems with this hunk:

>
>
> diff -d -urpN libunwind.0/include/libunwind_i.h
> libunwind.1/include/libunwind_i.h
> --- libunwind.0/include/libunwind_i.h   2012-02-13 18:43:50.000000000 +0100
> +++ libunwind.1/include/libunwind_i.h   2012-02-21 17:56:59.578536682 +0100
> @@ -303,6 +303,32 @@ struct elf_image
>     size_t size;               /* (file-) size of the image */
>   };
>
> +struct elf_dyn_info
> +  {
> +    struct elf_image ei;
> +    unw_dyn_info_t di_cache;
> +    unw_dyn_info_t di_debug;    /* additional table info for .debug_frame
> */
> +#if UNW_TARGET_IA64
> +    unw_dyn_info_t ktab;
> +#endif
> +#if UNW_TARGET_ARM
> +    unw_dyn_info_t di_arm;      /* additional table info for .ARM.exidx */
> +#endif
> +  };
> +
> +static void inline invalidate_edi (struct elf_dyn_info *edi)
> +{
> +  if (edi->ei.image)
> +    munmap (edi->ei.image, edi->ei.size);
> +  memset (edi, 0, sizeof (*edi));
> +  edi->di_cache.format = -1;
> +  edi->di_debug.format = -1;
> +#if UNW_TARGET_ARM
> +  edi->di_arm.format = -1;
> +#endif
> +}
> +
> +
>  /* Provide a place holder for architecture to override for fast access
>    to memory when known not to need to validate and know the access
>    will be local to the process. A suitable override will improve

Could you please test that the patches apply to HEAD or even better
point me to a git repo I can pull from.

Prefer example-core-unwind.c to be moved to the tests directory like I
had in this branch:

https://github.com/adsharma/libunwind/tree/coredump

Other than that, nothing jumped out at me. Another round of regression
testing on x86 and ARM and we should be done.

 -Arun