Allows launch config to override launch target connection type

[betamaxbandit]

When launching a GDB debug session using the Launch Bar and a Launch Target, now the launch takes into account a REMOTE_TCP value in the launch configuration before considering the type of the Launch Target.

Improves the change submitted in:
Bug 547881 - Allow targets to set ATTR_REMOTE_TCP

Fixes issue #622

[betamaxbandit]

Hi @jonahgraham ,

I'm not sure how to test this!
Either a manual UI test could be done or a JUnit added.
Manual test: I tried a manual test but didn't get past trying to launch it [1]

JUnit: I've investigated this a bit and it looks doable, but not sure whether it's worth it [2]

Do you have any advice please?

[1] Manual test
Using CDT New C/C++ Project wizard, created a CMake Project.

In Preferences > Core Build Toolchains, the auto detected GCC is shown:
Core Build Toolchains
Available Toolchains
Type Name OS Arch
GCC win32 x86_64: C:\msys64\mingw64\bin\gcc.exe win32 x86_64

In Preferences > CMake, I downloaded an example CMake toolchain file (MSYS2.cmake) from internet and added it to the toolchain above.

CMake
Toolchain Files
Toolchain File Toolchain
MSYS2.cmake win32 x86_64: C:\msys64\mingw64\bin\gcc.exe

I created a new Generic Target launch target:
Name: gen
Operating System: win32
CPU Architecture: x86_64

I built the project. An executable was successfully generated.

Click Launch Bar Run (Launch in Run mode) button
Console: No action specified. Skipping.

[2] Junit
I noticed org.eclipse.cdt.tests.dsf.gdb.launching.TestLaunchDelegate. I can extend this class and override

GdbLaunchDelegate.getLaunch(ILaunchConfiguration, String)

And then set the launch target using GdbLaunch.setLaunchTarget(ILaunchTarget)

Then setup a launch configuration which includes IGDBLaunchConfigurationConstants.ATTR_REMOTE_TCP attr set to true.

Then launch and inspect whether the FinalLaunchSequence.stepRemoteConnection(RequestMonitor) indeed attempts a isTcpConnection rather than a serial connect. [Not sure how to detect this]

[jonahgraham]

This change lgtm - I can merge it after I branch which is in the new few days.

My feeling for a test here is that a lot of scaffolding is probably needed to automate this test. I suspect the dev effort required to do it for this change doesn't warrant it. But I think we should make sure that at least a manual test runs, even if it is just for the default case where the attribute is not in the launch already. From your description it sounds like this has exposed other problems with the cmake integration. Some of them may already be reported in this query

[jonahgraham]

@betamaxbandit and I had a call about this and we verified that the behavior wasn't regressing as requested. So this is good to go. I am going to rebase the change to make sure there are no conflicts/issues

[jonahgraham]

Thanks @betamaxbandit for this fix.