Open ghost opened 5 years ago
I am observing a crash with your reproducer, but my crash stack does not look exactly the same as yours:
(gdb) bt
0 0x00007ffff7a89d77 in __libc_dlopen_mode () at /usr/local/google/home/chowski/dynamorio/core/lib/statsx.h:623
1 0x00007ffff7a72020 in __nss_lookup_function () at /usr/local/google/home/chowski/dynamorio/core/lib/statsx.h:188
2 0x00007ffff7a720fc in __nss_lookup () at /usr/local/google/home/chowski/dynamorio/core/lib/statsx.h:189
3 0x725f32656d616e79 in ?? ()
4 0x74736f6874656700 in ?? ()
5 0x5f33656d616e7962 in ?? ()
6 0x6e61637465670072 in ?? ()
7 0x725f656d616e6e6f in ?? ()
8 0x616e6e6f6e616300 in ?? ()
9 0x554e203d3d20656d in ?? ()
... hundreds, goes all the way to 972 frames
I'm not quite sure what the difference is, or why my nss_* functions are claiming to be in statsx.h
Xref the original discussion here: https://groups.google.com/forum/#!topic/DynamoRIO-Users/Sk4D0w2LC7s
Following up on this: the issue in both backtraces seem to be during dlopen
in libc
. Is there a good starting point for further investigation to help determine the cause (and eventually resovle) this error?
The following client produces a segmentation fault when run, due to some loading issue:
Here is the trace:
This is possibly related to static linking issues here: https://bugzilla.redhat.com/show_bug.cgi?id=89698 The solution described there (using an IP address directly instead of "localhost") seems to resolve the issue.