[Lldb-commits] [PATCH] D66994: [lldb][CMake] Fix windows-x86-debug compilation with python enabled using multi-target generator

Mon Sep 2 03:00:14 PDT 2019

leonid.mashinskiy added a comment.

Since this patch appeared to be arguable from your point of view I will write some comments before making any further changes.

In D66994#1652868 <https://reviews.llvm.org/D66994#1652868>, @amccarth wrote:

> - I use the VS project only for code browsing and debugging.  For building and tests, I use the ninja project.  So I'm not very familiar with the problem you're trying to fix.

Since there are lot of tablegenned code just opening project in VS without building it makes lot's of code unresolved for producivity tools providing smart code navigation, highlighting and etc (like Intellisence or Resharper) which helps me understand code faster.
Also I find it useful to be able to compile single cpp file from VS.
And also building from VS was working on 9 branch so I just thought it was broken unintentionally and can be fixed.

In D66994#1652868 <https://reviews.llvm.org/D66994#1652868>, @amccarth wrote:

> - Some of these changes you're undoing caused a lot of pain when they were first put in, so let's be really sure if we need to undo this.

I am not just undoing latest changes, that's why there are so many copy-pasted code. If CMAKE_BUILD_TYPE is set (which is set when building using ninja or other single-target generator) then logic should flow the same way as without my changes. I'm just supporting third way when CMAKE_BUILD_TYPE is not selected and configuration is deduced on built time and paths to selected libraries are selected using CMake expression generators as it was in 9 branch.

In D66994#1652868 <https://reviews.llvm.org/D66994#1652868>, @amccarth wrote:

> - lldb tests don't work with the debug Python interpreter, which is why I've been complaining about test failures approximately forever when the bots seem fine.

I found it easier to debug some tests in full debug environment, though I didn't try to run all test suite in debug.
Also we found some tests to be flaky even in release configurations but bot had never showed that.

If you are ok with my explanations I will try to extract some code to reduce copy-pasting, otherwise this patch can be abandoned.

Also I have a question about x86 windows configuration - is it officially supported or it will be abandoned soon? Because now it is completely broken.

Repository:
  rLLDB LLDB

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D66994/new/

https://reviews.llvm.org/D66994