Community
Participate
Working Groups
With bug 347245 we now have a preference to default the value for non-stop mode. If I turn that preference to enabled, and launch a hardware debug session with GDB <= 6.8, the launch will abort with: !MESSAGE Non-stop mode is only supported starting with GDB 6.8.50 The problem is that the hardware launch delegate re-uses GdbLaunchDelegate, which checks if non-stop is allowed, and that the hardware launch does not have any UI to enable/disable non-stop. Therefore, we use the preference value to know if non-stop should be on, and things fail. For non-hardware launches, it is up to the user to disable non-stop for any launch using GDB <= 6.8. But for those launches there is a UI.
The workaround for now is to make the preference for non-stop to be disabled in this case of an older GDB for hardware debug
(In reply to comment #1) > The workaround for now is to make the preference for non-stop to be disabled in > this case of an older GDB for hardware debug I have the global preference for non-stop mode disabled (unchecked) and I can confirm that non-stop mode is not enabled in GDBJtagDSFFinalLaunchSequence. However, the bad behaviour reported in bug 348043 when using GDB 6.8.x is still observed.
Same problem for post-mortem launches of older GDBs
Created attachment 201017 [details] Fix that disables non-stop automatically for post-mortem and hardware launches Now that we have a preference that can enable non-stop automatically, this patch disables non-stop in the case of post-mortem and hardware launches. For those launches, non-stop is not applicable and cannot be disabled by the user. This avoid running into the problem of a debug session with GDB 6.8 failing to start because it does not support non-stop.
I'll commit the patch to both master and cdt_8_0
Created attachment 201018 [details] Patch for cdt_8_0 When testing, I realized that cdt_8_0 was a little different and needed some extra changes. The problem was that fIsNonStopSession was being set in getLaunch() which is called before preLaunchCheck(). Therefore, we hadn't force non-stop to off yet, and things would abort again. In Bug 339550, I moved the use of fIsNonStopSession to much later, which is what I did in this patch.
Sergey can you review?
In GdbLaunchDelegate.java and GDBJtagDSFLaunchConfigurationDelegate.java: turn if off -> turn it off
*** cdt git genie on behalf of Marc Khouzam *** Bug 348091: Cannot launch hardware or post-mortem debugging for older GDBs when preference for non-stop is enabled [*] http://git.eclipse.org/c/cdt/org.eclipse.cdt.git/commit/?id=db62a47f3e60cca608b8c7bbd5a00afd27805086
*** cdt git genie on behalf of Marc Khouzam *** Bug 348091: Cannot launch hardware or post-mortem debugging for older GDBs when preference for non-stop is enabled [*] http://git.eclipse.org/c/cdt/org.eclipse.cdt.git/commit/?id=e1505b889963d9dc944ded4f26ae9390d639c712
(In reply to comment #8) > In GdbLaunchDelegate.java and GDBJtagDSFLaunchConfigurationDelegate.java: > turn if off -> turn it off Thanks, I changed it.
