Re: building/using address-sanitizer-enabled emacs?

[Jim Meyering]

On Tue, May 9, 2017 at 8:18 AM, Eli Zaretskii <address@hidden> wrote:
>> From: Jim Meyering <address@hidden>
>> Date: Mon, 8 May 2017 22:48:06 -0700
>> Cc: Paul Eggert <address@hidden>, emacs-devel <address@hidden>
>>
>>   echo foo |gpg -c > foo.gpg
>>   src/temacs -q foo.gpg 2> err
>>
>> which reported this stack buffer overrun:
>>
>> ==24522==ERROR: AddressSanitizer: stack-buffer-overflow on address
>> 0x7ffd5cc4d928 at pc 0x00000073fd76 bp 0x7ffd5cc4d910 sp
>> 0x7ffd5cc4d908
>> READ of size 8 at 0x7ffd5cc4d928 thread T0
>>     #0 0x73fd75 in PSEUDOVECTORP /home/j/w/co/emacs/src/lisp.h:1454
>>     #1 0x7438c9 in BUFFERP /home/j/w/co/emacs/src/buffer.h:887
>>     #2 0x9c64b6 in call_process /home/j/w/co/emacs/src/callproc.c:702
>>     #3 0x9c41db in Fcall_process /home/j/w/co/emacs/src/callproc.c:270
>>     #4 0x8e3122 in funcall_subr /home/j/w/co/emacs/src/eval.c:2799
>>     #5 0x8e2aa1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2744
>>     #6 0x8e072d in Fapply /home/j/w/co/emacs/src/eval.c:2375
>>     #7 0x8e3122 in funcall_subr /home/j/w/co/emacs/src/eval.c:2799
>>     #8 0x8e2aa1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2744
>>     #9 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #10 0x8e4b55 in funcall_lambda /home/j/w/co/emacs/src/eval.c:3022
>>     #11 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #12 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #13 0x8e4b55 in funcall_lambda /home/j/w/co/emacs/src/eval.c:3022
>>     #14 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #15 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #16 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #17 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #18 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #19 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #20 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #21 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #22 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #23 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #24 0x8e072d in Fapply /home/j/w/co/emacs/src/eval.c:2375
>>     #25 0x8e3122 in funcall_subr /home/j/w/co/emacs/src/eval.c:2799
>>     #26 0x8e2aa1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2744
>>     #27 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #28 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #29 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #30 0x8e1f8d in call6 /home/j/w/co/emacs/src/eval.c:2649
>>     #31 0x8034eb in Finsert_file_contents 
>> /home/j/w/co/emacs/src/fileio.c:3602
>>     #32 0x8e367f in funcall_subr /home/j/w/co/emacs/src/eval.c:2831
>>     #33 0x8e2aa1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2744
>>     #34 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #35 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #36 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #37 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #38 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #39 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #40 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #41 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #42 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #43 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #44 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #45 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #46 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #47 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #48 0x8e2ae1 in Ffuncall /home/j/w/co/emacs/src/eval.c:2746
>>     #49 0x98c89b in exec_byte_code /home/j/w/co/emacs/src/bytecode.c:641
>>     #50 0x8e4604 in funcall_lambda /home/j/w/co/emacs/src/eval.c:2944
>>     #51 0x8e3f96 in apply_lambda /home/j/w/co/emacs/src/eval.c:2881
>>     #52 0x8df8ab in eval_sub /home/j/w/co/emacs/src/eval.c:2265
>>     #53 0x8dda2c in Feval /home/j/w/co/emacs/src/eval.c:2042
>>     #54 0x8df175 in eval_sub /home/j/w/co/emacs/src/eval.c:2223
>>     #55 0x945574 in readevalloop /home/j/w/co/emacs/src/lread.c:1947
>>     #56 0x9425f5 in Fload /home/j/w/co/emacs/src/lread.c:1352
>>     #57 0x8df41e in eval_sub /home/j/w/co/emacs/src/eval.c:2234
>>     #58 0x8dda2c in Feval /home/j/w/co/emacs/src/eval.c:2042
>>     #59 0x751a34 in top_level_2 /home/j/w/co/emacs/src/keyboard.c:1121
>>     #60 0x8d9c05 in internal_condition_case 
>> /home/j/w/co/emacs/src/eval.c:1326
>>     #61 0x751a97 in top_level_1 /home/j/w/co/emacs/src/keyboard.c:1129
>>     #62 0x8d83e9 in internal_catch /home/j/w/co/emacs/src/eval.c:1091
>>     #63 0x751899 in command_loop /home/j/w/co/emacs/src/keyboard.c:1090
>>     #64 0x75033f in recursive_edit_1 /home/j/w/co/emacs/src/keyboard.c:697
>>     #65 0x7506dd in Frecursive_edit /home/j/w/co/emacs/src/keyboard.c:768
>>     #66 0x74bbb9 in main /home/j/w/co/emacs/src/emacs.c:1687
>>     #67 0x7f40c7732400 in __libc_start_main (/lib64/libc.so.6+0x20400)
>>     #68 0x40d369 in _start (/home/j/w/co/emacs/src/temacs+0x40d369)
>>
>> Address 0x7ffd5cc4d928 is located in stack of thread T0 at offset 136 in 
>> frame
>>     #0 0x9c8c15 in child_setup /home/j/w/co/emacs/src/callproc.c:1179
>>
>>   This frame has 2 object(s):
>>     [32, 40) 'display'
>>     [96, 104) 'tmp' <== Memory access at offset 136 overflows this variable
>> HINT: this may be a false positive if your program uses some custom
>> stack unwind mechanism or swapcontext
>
> I admit that I don't understand this report.  At the point where the
> report claims there was buffer overflow, child_setup is not in the
> call stack, because it is/was called in another process after vfork:
> the callstack shows the stack of the Emacs process, whereas
> child_setup is called by a child process.  So either the report shows
> a stack of a wrong process, or something else is going on.  Or maybe I
> simply don't understand how to read this report.

Thanks for looking.
I confess I have not done so. Does this caveat apply?

  HINT: this may be a false positive if your program uses some custom
  stack unwind mechanism or swapcontext