[Bug default/29637] New: dzw fails with Unknown DWARF DW_OP_255

[Bug default/29637] New: dzw fails with Unknown DWARF DW_OP_255

[nolange79 at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29637

            Bug ID: 29637
           Summary: dzw fails with Unknown DWARF DW_OP_255
           Product: dwz
           Version: unspecified
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: default
          Assignee: nobody at sourceware dot org
          Reporter: nolange79 at gmail dot com
                CC: dwz at sourceware dot org
  Target Milestone: ---

Created attachment 14370
  --> https://sourceware.org/bugzilla/attachment.cgi?id=14370&action=edit
binary reproduzing the bug

dwz fails to process the attached file with 'Unknown DWARF DW_OP_255'

dwz version 0.14

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/29637] dzw fails with Unknown DWARF DW_OP_255

[mliska at suse dot cz]

https://sourceware.org/bugzilla/show_bug.cgi?id=29637

Martin Liska <mliska at suse dot cz> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mliska at suse dot cz
             Status|UNCONFIRMED                 |WAITING
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2022-09-29

--- Comment #1 from Martin Liska <mliska at suse dot cz> ---
Confirmed. How did you build the library, can you please share build log?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/29637] dzw fails with Unknown DWARF DW_OP_255

[jakub at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29637

Jakub Jelinek <jakub at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at redhat dot com

--- Comment #2 from Jakub Jelinek <jakub at redhat dot com> ---
This is while processing of
 <10><16d34>: Abbrev Number: 12 (DW_TAG_variable)
    <16d35>   DW_AT_abstract_origin: <0x63976>
    <16d39>   DW_AT_location    : 0x29d51 (location list)
    <16d3d>   DW_AT_GNU_locviews: 0x29d47
and seems even readelf -wo shows total garbage for that:
    00029d51 v000000000000001 v000000000000000 views at 00029d47 for:
             000000000000af34 000000000000b10b (DW_OP_implicit_value 16 byte
block: ff ff ff ff ff ff ff ff 30 b1 0 0 0 0 0 0 )
    00029d75 v000000000000000 v000000000000000 views at 00029d49 for:
             000000000000b161 ffffffff109e0012 ((User defined location op
0xff))
    00039d86 v000000000000000 v000000000000000 views at 00029d4b for:
             e201000000000000 e2000000000000cb ((Unknown location op 0x0))
(start > end)
    00039e63 v000000000000000 v000000000000000 views at 00029d4d for:
             d64e9f7b80910004 d686000000000000 ()
    00039e75 v000000000000000 v000000000000000 views at 00029d4f for:
             8091000400000000 0000000000009f7b () (start > end)
    00039e87 v000000000000034 v00000000000002f views at 00029d51 for:
             0000000000000000 0000cc1100000001 ()
    00039e99 v000000000000000 v000000000000000 views at 00029d54 for:
             00000000cc480000 00d64e5e00010000 ()
    00039eab v000000000000000 v000000000000000 views at 00029d56 for:
             000000d686000000 00005e0001000000 ()
    00039ebd v000000000000000 v00000000000000b views at 00029d58 for:
             0000000000000000 0000000100000000 ((Unknown location op 0x0))
    00046ae0 v000000000000031 v000000000000000 views at 00029d5a for:
             0000000000000000 0000436000000000 ()
    00046af2 v000000000000000 v000000000000000 views at 00029d5d for:
             0000000043690000 0043695500010000 ()
    00046b04 v000000000000000 v000000000000000 views at 00029d5f for:
             000000436e000000 436e560001000000 ()
    00046b16 v000000000000012 v000000000000000 views at 00029d61 for:
             0000437300000000 7355000100000000 ((Unknown location op 0x0))
    00046b6b v00000000000081e v30ffffffffffffff views at 00029d63 for:
             00000000004fe100 00004fe154000100 ()
    00046b7d v000000000000031 v000000000000000 views at 00029d6e for:
             000000004fe20000 9f5501f300040000 ()
    00046b8f <End of list>

Bet GCC emitted there DW_OP_implicit_value 16 with just 8 bytes of data after
it.
I vaguely remember fixing some issues like this, I think e.g.
https://gcc.gnu.org/PR99562 and
https://gcc.gnu.org/PR103046
Anyway, doesn't seem like a dwz bug to me.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/29637] dzw fails with Unknown DWARF DW_OP_255

[yumeyao at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29637

YumeYao <yumeyao at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |yumeyao at gmail dot com

--- Comment #3 from YumeYao <yumeyao at gmail dot com> ---
I can confirm this is a gcc bug. The only reason one can blame dwz for is dwz
is not robust enough to handle broken DWARF data.

fmtlib 9.0~ with gcc 6.0~10.5 triggers this issue. gcc 5.5 and 11.0 works fine.

I have backported gcc PR99562 and PR103046 to gcc 6.5 and gcc 7.5 and the issue
is gone. Specifically, PR99562 is the only one needed to fix the 'broken' DWARF
data, wheres PR103046 seems to fix the 'wrong' (but not broken) data.

What is remaining unclear is why gcc 5.5 (which has PR66728 merged, which
according to @Jakub, is the cause of the issue) doesn't trigger this issue on
fmtlib 9.0~.

That said, still, is there any chance that dwz can be more tolerant to broken
dwz? or maybe we can give dwz a switch to discard broken sections with warning
not error?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/29637] dzw fails with Unknown DWARF DW_OP_255

[jakub at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29637

Jakub Jelinek <jakub at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |INVALID
             Status|WAITING                     |RESOLVED

--- Comment #4 from Jakub Jelinek <jakub at redhat dot com> ---
dwz does the right thing, whenever it sees something unknown, it fails to
change the DWARF in the binary or shared library with such a bug.  Anything
else would be extremely dangerous.

-- 
You are receiving this mail because:
You are on the CC list for the bug.