[Lldb-commits] [PATCH] D32930: New framework for lldb client-server communication tests.

[Pavel Labath via Phabricator via lldb-commits]

labath added a comment.

I am going to commit this, so we can move forward with the remote testing aspect.

I'd like to reiterate that we are not planning to remove any existing tests until both remote and debugserver testing work. At that point, we'd like to start removing the existing tests, but we will bring that up again when the time comes. We can also certainly allow for some temporal flexibility if that happens to fall into someone release deadline or such.



================
Comment at: unittests/tools/CMakeLists.txt:1
+if(UNIX AND NOT APPLE)
+  add_subdirectory(lldb-server)
----------------
beanz wrote:
> labath wrote:
> > beanz wrote:
> > > labath wrote:
> > > > This is not what I meant. The only targets (at least until we have debugserver support) that can realistically pass these tests are linux, android, and netbsd. The other targets (right now, I guess that would mean freebsd) don't even pretend to support debugging via lldb-server, so we should not fail their build because of that. Check for usages of CMAKE_SYSTEM_NAME to see how to discriminate those.
> > > Darwin pretends to support lldb-server in several places, it would be nice to be able to run these tests on Darwin if they work. One of my big goals for the future of testing on LLDB is to get to the point where the only differences in test coverage when running tests on different hosts is truly platform-specific code. Today we are nowhere near that.
> > > 
> > > Also, as Pavel pointed out in email, the lldb-server tests are also run against debugserver, so we need to make sure that still works too.
> > Which lldb-server support do you refer to here?
> > 
> > There is some llgs (debugging) support in lldb-server, but I have no idea what's the state of it -- it was added by Todd during his week of code as an "NFC" commit, and it hasn't been touched since. I'd like to avoid this keeping the build red if there is no intention of working on it. 
> > 
> > The "platform" mode of lldb-server should work on darwin afaik, and we definitely want to be able to run it there. It's not what we are focusing on now though. We'd like to migrate the "debug" tests first (there are no "platform" tests), so the old ones can be removed. 
> > 
> > In any case, I think of the apple exclusion part as a temporary thing, so we can check this in without breaking the build, we will pretty soon want to include it as well, so that we can run debugserver tests, at least. (At which point we will need a different way of disabling unsupported tests).
> I don't know the state of this stuff on Darwin either. I had spent a little time a few weeks ago trying to get lldb to use lldb-server on Darwin to see if I could answer that question, but I didn't get very far before I had to stop.
> 
> As long as the Apple exclusion is temporary I'm fine with this as-is. I'll see if I can free up some time this summer to look more closely at lldb-server on Darwin and figure out what state it is in.
Yep, it's definitely temporary. Getting this thing working with debugserver is my #2 priority for this project (mainly because I expect it to be easy). Right now we are looking at the remote testing aspect of this, which is going to be more tricky, and without which we are dead in the water. If you have some thoughts or requirements on this aspect, now's the time to speak :). I am planning to make this generic enough so it can use both ssh and adb as a conduit, at which point I hope it can be used with iOS as well.

As for lldb-server on darwin, that will have to come from you guys, as I believe there is still a lot of work to be done to make it functional.


https://reviews.llvm.org/D32930