Re: [PATCHv2] support ZSTD compression algorithm

["Martin Liška" <mliska@suse.cz>]

On 12/19/22 16:09, Mark Wielaard wrote:
> Hi Martin,
> 
> On Mon, 2022-12-19 at 15:19 +0100, Martin Liška wrote:
>> On 12/15/22 14:17, Mark Wielaard wrote:
>>> Is there a particular way you are running eu-readelf? Is it with
>>> generic -w or -a, or decoding a specific section type?
>>
>> Hello.
>>
>> $ LD_LIBRARY_PATH=./libelf ./src/readelf -w
>> ~/Programming/testcases/a.out
>>
>> where I get:
>>
>> ./src/readelf: cannot get debug context descriptor: No DWARF
>> information found
>>
>> DWARF section [37] '.debug_info' at offset 0x1ab2:
>>   [Offset]
>> ./src/readelf: cannot get next unit: no error
>>
>> Call frame information section [13] '.eh_frame' at offset 0x4a8:
>> ...
>>                                                                      
>>                                                                      
>>       t��o5��=I�iAp@a����S^R/<�����^�qi�ַp@
> 
> [...]
> 
>> So basically a garbage. And I don't know how to bail out properly?
> 
> Aha. If you have that a.out somewhere I can take a look. I suspect this
> is because we expect all .debug sections to have been decompressed in
> libdw/dwarf_begin_elf.c, but that isn't really true, see check_section
> in that file which has:
> 
>   if ((shdr->sh_flags & SHF_COMPRESSED) != 0)
>     {
>       if (elf_compress (scn, 0, 0) < 0)
>         {
>           /* It would be nice if we could fail with a specific error.
>              But we don't know if this was an essential section or not.
>              So just continue for now. See also valid_p().  */
>           return result;
>         }
>     }

Sure, there's a file example:
https://splichal.eu/tmp/zstd.out

$ ./src/readelf -Sz zstd.out
...
[35] .debug_abbrev        PROGBITS     0000000000000000 000018b4 00000177  0 C      0   0  1
     [ELF ZSTD (2) 00000318  1]

$ LD_LIBRARY_PATH=libelf/ ./src/readelf -w zstd.out
(prints garbage if not configured with zstd)

Please use code from my branch:
https://sourceware.org/git/?p=elfutils.git;a=shortlog;h=refs/heads/users/marxin/try-zstd-support-v2

Cheers,
Martin

> 
> We should probably adjust valid_p so it produces a more appropriate
> error message and/or add additional checks in readlelf.c.
> 
> But lets do that after this patch goes in.
> 
> Cheers,
> 
> Mark