[Bug binutils/32927] New: readelf --dwarf-check --debug-dump=Ranges produces nonsense warnings for DWARF 4 and DWARF5
https://sourceware.org/bugzilla/show_bug.cgi?id=32927 Bug ID: 32927 Summary: readelf --dwarf-check --debug-dump=Ranges produces nonsense warnings for DWARF 4 and DWARF5 Product: binutils Version: 2.45 (HEAD) Status: UNCONFIRMED Severity: normal Priority: P2 Component: binutils Assignee: unassigned at sourceware dot org Reporter: hmeyer.eu at gmail dot com Target Milestone: --- example output from --dwarf-check --debug-dump=Ranges: DWARF4/debug_ranges: Offset BeginEnd 36d4 36d4 (start == end) 36d8 36dc readelf: Warning: There is a hole [0 - 0x30] in .debug_ranges section. 0030 36d4 36d8 0030 36dc 36e5 0030 readelf: Warning: There is a hole [0x30 - 0x60] in .debug_ranges section. 0060 3704 3704 (start == end) 0060 3708 370c 0060 readelf: Warning: There is a hole [0x60 - 0x90] in .debug_ranges section. 0090 3704 3708 0090 370c 3715 0090 DWARF5/debug_rnglists: Offset BeginEnd 000c 36d4 (base address) 0015 36d4 36d4 (start == end) 0018 36d8 36dc 001b readelf: Warning: There is a hole [0xc - 0x1c] in .debug_rnglists section. 001c 36d4 (base address) 0025 36d4 36d8 0028 36dc 36e5 002b readelf: Warning: There is a hole [0x1c - 0x2c] in .debug_rnglists section. 002c 3704 (base address) 0035 3704 3704 (start == end) 0038 3708 370c 003b The hole check always triggers, even if the other output shows that there are no holes. If the hole check triggers, it prevents the overlap check from triggering, even if there are overlaps. However, it is normal for compilers (GCC 13 and later) to produce overlapping rangelists in debug_rnglists as noted in the discussion of bug 30792. readelf --dwarf-check --debug-dump=Ranges seems to be broken for all versions since 2.35, which is the oldest I tested (debian oldstable). I have not attached a binary, it seems to be wrong for all inputs. -- You are receiving this mail because: You are on the CC list for the bug.
[Bug binutils/32927] readelf --dwarf-check --debug-dump=Ranges produces nonsense warnings for DWARF 4 and DWARF5
https://sourceware.org/bugzilla/show_bug.cgi?id=32927
--- Comment #2 from Henning Meyer ---
The following C++ program represents the simplest example I could find that is
complicated enough to warrant the creation of discontinuous ranges in debug
information:
#include
struct Base {};
void func() {
static std::unique_ptr varStatic = std::make_unique();
}
int main(int, char**) {
func();
return 0;
}
I compile with g++-14 -g3 -gdwarf-5 -O2 -c:
readelf --dwarf-check --debug-dump=Ranges
Contents of the .debug_rnglists section:
readelf: Warning: Range lists in .debug_rnglists section start at 0xc
Table at Offset: 0:
Length: 0x5d
DWARF version: 5
Address size:8
Segment size:0
Offset entries: 0
Offset BeginEnd
000c 0024 (base address)
0015 0024 002e
0018 003c 0043
001b
readelf: Warning: There is a hole [0xc - 0x1c] in .debug_rnglists section.
001c 002e (base address)
0025 002e 002e (start == end)
0028 003c 0043
002b
readelf: Warning: There is a hole [0x1c - 0x2c] in .debug_rnglists section.
002c 002e (base address)
0035 002e 003c
0038 0043 004f
003b 0050 0053
003e 0055 005e
0041
readelf: Warning: There is a hole [0x2c - 0x42] in .debug_rnglists section.
0042 0076
004c 0017
0056 000e
0060
(That is the complete output). I can attach the binary, it is 180K in size.
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug binutils/32927] readelf --dwarf-check --debug-dump=Ranges produces nonsense warnings for DWARF 4 and DWARF5
https://sourceware.org/bugzilla/show_bug.cgi?id=32927 --- Comment #4 from Henning Meyer --- I've applied the patch, it removes the bogus warnings on my test set. -- You are receiving this mail because: You are on the CC list for the bug.
