https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86108
Bug ID: 86108
Summary: crash during unwinding with -O2
Product: gcc
Version: 8.1.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: guillaume at morinfr dot org
Target Milestone: ---
Created attachment 44259
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=44259&action=edit
reproducer
I have a program that crashes very reliably after an exception is thrown. The
code source is attached. It's not as small as I would like but trying to
simplify it significantly makes the crash go away. It crashes very early in
main() though.
Compile with "-std=c++17 -O2" on a GNU/Linux amd64 box, then run "-f RYz1q" to
reproduce.
Starting program: ./repro -f RYz1q
Catchpoint 1 (exception thrown), 0x00007ffff7aeaa4d in __cxa_throw () from
/usr/lib/x86_64-linux-gnu/libstdc++.so.6
(gdb) bt
#0 0x00007ffff7aeaa4d in __cxa_throw () from
/usr/lib/x86_64-linux-gnu/libstdc++.so.6
#1 0x0000000000406371 in RR::RR (this=<optimized out>, _readSize=<optimized
out>, maker=<optimized out>, headerCopy=<optimized out>, start=..., end=...,
_writer=0x0) at repro.cc:1370
#2 0x0000000000402353 in main () at repro.cc:1550
#3 0x00007ffff71bf2e1 in __libc_start_main (main=0x401bc0 <main>, argc=3,
argv=0x7fffffffe018, init=<optimized out>, fini=<optimized out>,
rtld_fini=<optimized out>, stack_end=0x7fffffffe008) at ../csu/libc-start.c:291
#4 0x0000000000402a1a in _start () at repro.cc:1332
(gdb) c
Continuing.
Program received signal SIGSEGV, Segmentation fault.
0x00007ffff754cdaf in _Unwind_Resume () from
/lib/x86_64-linux-gnu/libgcc_s.so.1
(gdb) bt
#0 0x00007ffff754cdaf in _Unwind_Resume () from
/lib/x86_64-linux-gnu/libgcc_s.so.1
#1 0x000000000040185c in main (argc=<optimized out>, argv=<optimized out>) at
repro.cc:1623
#2 0x00007ffff71bf2e1 in __libc_start_main (main=0x401bc0 <main>, argc=3,
argv=0x7fffffffe018, init=<optimized out>, fini=<optimized out>,
rtld_fini=<optimized out>, stack_end=0x7fffffffe008) at ../csu/libc-start.c:291
#3 0x0000000000402a1a in _start () at repro.cc:1332
It seems to be a missing null pointer check:
==22653== Command: ./repro -f RYz1q
==22653==
==22653== Invalid read of size 8
==22653== at 0x54CDDAF: _Unwind_Resume (in
/lib/x86_64-linux-gnu/libgcc_s.so.1)
==22653== by 0x40185B: main.cold.131 (repro.cc:1543)
==22653== by 0x56F52E0: (below main) (libc-start.c:291)
==22653== Address 0x40 is not stack'd, malloc'd or (recently) free'd
Using g++ 7.x, -O1 or adding "-fno-reorder-blocks-and-partition" makes the code
work (the exception is caught and the program prints "CAUGHT EXCEPTION - IT
WORKED").
