Fred Smith wrote:
> I apologize up front, I'm sure this isn't the right place to ask, but
> I'm not sure where I should ask, instead. Hoping for some helpful hints
> in that regard...
>
> I'm using (trying to use...) the -fsanitize=thread feature. Though I'm
> finding the diagnostics rather cryptic, that isn't my issue here.
> Suddenly, a few days ago, I started getting this error while running the
> app i'm testing:
<snip>
> [New Thread 0x7fcd2f9ea700 (LWP 8152)]
> [New Thread 0x7fcd2f1e9700 (LWP 8153)]
> ThreadSanitizer: longjmp() is not supported
<snip>
> [Thread 0x7fcd47858700 (LWP 8141) exited]
>
> The message about longjmp() is the issue. This app doesn't use longjmp
> or siglongjump or any other of its cousins. It's possible that some
> 3rd-party lib (Oracle??) might, but I don't know why it has been working
for several
> weeks and suddenly starts giving me this error...
>
> Clues welcome, thanks!
Many years ago, a couple young consultants came to me where we were
working (I was the "senior technical resource"), and told me the program
kept crashing in a library. We ran it under the debugger, and we saw the
library call. We reran it, with me at the keyboard, and I did something
they had no idea about: I stepped *into* the call, (the old Sun debugger,
stepi, rather than step, or stepn). I continued to step, working my way
down, finding the calls it crashed in. Finally, I found the function that
was crashing, stepped into it - turned out to be a BEA Tuxedo call - and
then I, who had the authority, could call them, and even though it was
stripped... I could tell them the function it failed in, and more-or-less
what it seemed to be doing. 'Bout a day later, I got a message back from
the developers what the issue was.
You might try finding your problem that way; then you might have a clue as
to how to resolve it.
Oh, and if it's Oracle, don't forget to set TWO_TASK, I think it was
called (haven't used Pro*C more than a dozen years....)
mark