gdb/testsuite: resume-exception.exp - handle failure to hit breakpoint - #238
Conversation
|
Hi @akondrat-amd. Do you know why such a thing would happen? Is it a timeout issue, a specific architecture that behaves differently? A bug? |
It's an issue with rocjitsu that will be eventually fixed. |
fb3ecfd to
08d6037
Compare
If GDB fails to hit the breakpoint in raise_fpe, gpu_thread is never set and referencing it later would cause a Tcl error. Initialize gpu_thread to "invalid" before running and return early if it was not updated by the breakpoint handler.
08d6037 to
e80e2dc
Compare
Why would we need to adjust ROCgdb if it's a rocjitsu issue? |
It's our bug, we do not initialize the variable, rocjitsu just exposed it by failing in the right place. It's more cosmetic than functional. If the subtest fails, we get a TCL error with a call stack instead of test failure. |
Oops. Sorry, that wasn't clear to me. |
lumachad
left a comment
lumachad
left a comment
There was a problem hiding this comment.
This looks OK to me. Thanks.
4 participants
If GDB fails to hit the breakpoint in raise_fpe, gpu_thread is never
set and referencing it later would cause a Tcl error. Initialize
gpu_thread to "invalid" before running and return early if it was not
updated by the breakpoint handler.
Jira: AIROCGDB-640