Robert O'Callahan
Robert O'Callahan
Not easily reproduced locally :-/.
20 concurrent test runs on Anubis don't fail either.
I think that's `EBADF` which is a bit strange. You'd have to dig into rr to figure out which fd we tried to map in the tracee and whether that's...
You'll want to build from source first as Kyle suggested. After that, reproduce the bug, and when rr pauses as before, attach gdb to it and get a stack trace....
I don't see how it could be -75. `AutoRemoteSyscalls::check_syscall_result` should be called by `infallible_syscall(syscall_number_for_openat(arch()))`, and `check_syscall_result` should trigger an assertion failure for ret == -75. Can you figure out why...
> a long ret value of 4294967221 which is more than 2^32 It's actually less but I see your point. I guess on a 32-bit architecture we should be treating...
So I think we need to know the parameters to `AutoRemoteSyscalls::finish_direct_mmap`, in particular `backing_offset_pages` and `length` in `AutoRemoteSyscalls::finish_direct_mmap`. Sounds like that's overflowing and I'm not sure why... `backing_file_name` and the...
I cannot explain why we'd get EOVERFLOW with those parameters. "the number of pages used for length plus number of pages used for offset" would only be 455916345/4096 which is...
A simple attempt to reproduce this locally didn't work for me.