Cuda-gdb error on cuda12.6 while using python3.12-tui (original) (raw)

September 10, 2024, 6:57pm 1

While using cuda-gdb, I get the following error when my program( the one I am trying to debug )stops due to Illegal Warp Address:
(NOTE : this happens after my programs exits due to some error)
cuda-gdb has received a SIGSEGV and will attempt to get its own backtrace.

…thon3.12-tui| segv_handler() +0x4a
…ib/libc.so.6| ???
…ibexpat.so.1| XML_SetHashSalt() +0x13
…linux-gnu.so| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_Vectorcall() +0x5d
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0xb71
…n3.12.so.1.0| _PyObject_FastCallDictTstate() +0x1e6
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyObject_MakeTpCall() +0x2a8
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0xb71
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_Vectorcall() +0x5d
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0xb71
…n3.12.so.1.0| PyEval_EvalCode() +0xb5
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x64dd
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_CallMethodObjArgs() +0x10b
…n3.12.so.1.0| PyImport_ImportModuleLevelObject() +0x939
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x7139
…n3.12.so.1.0| PyEval_EvalCode() +0xb5
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x64dd
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_CallMethodObjArgs() +0x10b
…n3.12.so.1.0| PyImport_ImportModuleLevelObject() +0x939
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x7139
…n3.12.so.1.0| PyEval_EvalCode() +0xb5
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x64dd
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_CallMethodObjArgs() +0x10b
…n3.12.so.1.0| PyImport_ImportModuleLevelObject() +0x939
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x7139
…n3.12.so.1.0| PyEval_EvalCode() +0xb5
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x64dd
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_CallMethodObjArgs() +0x10b
…n3.12.so.1.0| PyImport_ImportModuleLevelObject() +0x939
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x7139
…n3.12.so.1.0| PyEval_EvalCode() +0xb5
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x64dd
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_CallMethodObjArgs() +0x10b
…n3.12.so.1.0| PyImport_ImportModuleLevelObject() +0x939
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x64dd
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_CallMethodObjArgs() +0x10b
…n3.12.so.1.0| PyImport_ImportModuleLevelObject() +0x939
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x7139
…n3.12.so.1.0| PyEval_EvalCode() +0xb5
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x64dd
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_CallMethodObjArgs() +0x10b
…n3.12.so.1.0| PyImport_ImportModuleLevelObject() +0x939
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x7139
…n3.12.so.1.0| PyEval_EvalCode() +0xb5
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x64dd
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_CallMethodObjArgs() +0x10b
…n3.12.so.1.0| PyImport_ImportModuleLevelObject() +0x939
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x7139
…n3.12.so.1.0| PyEval_EvalCode() +0xb5
…n3.12.so.1.0| ???
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x64dd
…n3.12.so.1.0| ???
…n3.12.so.1.0| _PyEval_EvalFrameDefault() +0x143b
…n3.12.so.1.0| ???
…n3.12.so.1.0| PyObject_CallFunctionObjArgs() +0xc1
…thon3.12-tui| gdbpy_colorize() +0x188
…thon3.12-tui| ext_lang_colorize() +0x53
…thon3.12-tui| source_cache::ensure() +0x444
…thon3.12-tui| source_cache::get_source_lines() +0x33
…thon3.12-tui| print_source_lines_base() +0x52f
…thon3.12-tui| print_frame_info() +0x207b
…thon3.12-tui| print_stack_frame() +0xa3
…thon3.12-tui| print_stop_event() +0x102
…thon3.12-tui| cli_base_on_normal_stop() +0xab
…thon3.12-tui| gdb::observers::observable<bpstat*, int>::notify() +0xc4
…thon3.12-tui| normal_stop() +0x2ff
…thon3.12-tui| fetch_inferior_event() +0x79c
…thon3.12-tui| gdb_wait_for_event() +0x53d
…thon3.12-tui| gdb_do_one_event() +0x13f
…thon3.12-tui| captured_command_loop() +0x3f

OS information: Arch Linux
version: cuda-12.6.68
driver: 560.35.3

I used pacman to install the toolkit. My program compiles fine but the debugger crashes.

steveu September 10, 2024, 7:57pm 2

Can you provide details on the host system you’re running on? What distribution of Linux, and its version? CUDA GDB has a number of distinct binaries that are build for various combinations of Python and TUI support, and the decision about which specific binary to use is made at runtime based upon distro properties.

mmason September 16, 2024, 9:33pm 3

A possible workaround while we investigate is to use “cuda-gdb-minimal” instead of “cuda-gdb”.
Could you give that a try and let us know if it works around the issue for you?

veraj Closed October 25, 2024, 10:05am 4

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.