The backtrace(3) function returns size_t which is unsigned:
backtrace(void **addrlist, size_t len);
And is documentet as returning:
The backtrace() function returns the number of elements that were filled
in the backtrace. [...]
The implementation in contrib/libexecinfo/unwind.c clearly knows that the return value can be all-ones:
if (ctx.n != (size_t)~0 && ctx.n > 0)
ctx.arr[--ctx.n] = NULL; /* Skip frame below __start */
This happens on a BeagleBoneBlack running
10.3-STABLE FreeBSD 10.3-STABLE #0 r300092
I am not sure what the proper fix is here, nor for that matter what the problem might be that causes it to return the all-ones value in the first place.
But at the very least ctx.n should be slammed to zero before returning it, so the code calling backtrace(3) doesn't run off the end of the world.
A commit references this bug:
Date: Mon Sep 21 16:43:39 UTC 2020
New revision: 365952
Pull in fix from upstream NetBSD rev. 1.5:
If Unwind_Backtrace is broken, ctx.n will still contain ~0, and we will
return that which poor behavior for the user, so return 0 instead.
We could document ~0 to be an error, but that would deviate from the
Linux behavior which is not desirable. Noted by Poul-Henning Kamp