[Bug gdb/22215] New: Triggered assert at dwarf2read.c:10986 elf_section_nr < dwp_file->num_sections

["alexander.v.shaposhnikov at gmail dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=22215

            Bug ID: 22215
           Summary: Triggered assert at dwarf2read.c:10986 elf_section_nr
                    < dwp_file->num_sections
           Product: gdb
           Version: HEAD
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: gdb
          Assignee: unassigned at sourceware dot org
          Reporter: alexander.v.shaposhnikov at gmail dot com
  Target Milestone: ---

This assert triggers if the section .strtab is the first one 
(probably, more generally (but i didn't verify that), the same issue can come
up if .strtab is not the last section.

Example: 

  main.cpp:
  struct Foo {
    int bar;
  };
  int main() {
    Foo foo;
    foo.bar = 1;
    int x = foo.bar + 1;
    return x;
  }

  build commands:
  g++ -fPIC -gsplit-dwarf -g -O0 main.cpp -o main.exe
  llvm-dwp main.dwo -o main.exe.dwp

  llvm-dwp (built from LLVM upstream, or, alternatively, 
  5.0 version should work as well) 
  places .strtab section at the beginning of the binary:
  Section Headers:
  [Nr] Name              Type             Address           Offset
       Size              EntSize          Flags  Link  Info  Align
  [ 0]                   NULL             0000000000000000  00000000
       0000000000000000  0000000000000000           0     0     0
  [ 1] .strtab           STRTAB           0000000000000000  000002a0
       0000000000000053  0000000000000000           0     0     1

  Next one needs to build GDB from the upstream + make sure the patch 
  https://sourceware.org/ml/gdb-patches/2017-09/msg00814.html is included
  (it fixes a separate issue, otherwise gdb will segfault much earlier).

  run gdb:
  gdb
  and see
  (gdb) file main.exe
Reading symbols from main.exe...dwarf2read.c:10986: internal-error: void
dwarf2_locate_common_dwp_sections(bfd*, asection*, void*): Assertion
`elf_section_nr < dwp_file->num_sections' failed.
A problem internal to GDB has been detected,

Some remarks: 
1. To workaround this issue one can use the following trick: 
objcopy --remove-section .strtab main.exe.dwp
objcopy will remove ".strtab" section but will create a new section ".shstrtab"
and will put it at the end. GDB works fine with the newly created .dwp.

2. It looks like that the value of
dwp_file->num_sections is not correct. I didn't dig deep into this issue, but
one potential scenario might look as follows (i'm not sure, it's just a guess,
might be completely wrong): 
  in the file elf.c, inside the function 
  bfd_section_from_shdr (bfd *abfd, unsigned int shindex),
  at the line 2242, case SHT_STRTAB: 
   for the section .strtab the function _bfd_elf_make_section_from_shdr is not
called 
   => the function bfd_section_init is not called => abfd->section_count is not
incremented.

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