[Lldb-commits] [PATCH] D61611: [JITLoaderGDB] Set eTypeJIT for objects read from JIT descriptors

Stefan Gränitz via Phabricator via lldb-commits lldb-commits at lists.llvm.org
Wed May 8 10:34:45 PDT 2019

sgraenitz marked 3 inline comments as done.
sgraenitz added a comment.

Thanks for your reply and thoughts about that.

> I'd strongly encourage you to try to come up with a testing strategy here.

Yes, I just added a simple lit test. What do you think?
I ran it in isolation on macOS (`UNSUPPORTED`) and on Ubuntu 18.04 (`PASS`) using:

  $ cd path/to/llvm-build
  $ ninja FileCheck llvm-config lli clang lldb
  $ python bin/llvm-lit -v /path/to/llvm-project/lldb/lit/Breakpoint/jitbp_elf.test

Comment at: lldb/lit/Breakpoint/Inputs/jitbp.cpp:2
+int jitbp() { return 0; }
+int main() { return jitbp(); }
lli already has a `main()` and so we have `jitbp()` here to set the breakpoint on. 

Comment at: lldb/lit/Breakpoint/jitbp_elf.test:1
+# REQUIRES: target-x86_64, system-linux, native
The test only works with ELF on Linux. Is the `REQUIRES` sufficient?

Comment at: lldb/lit/Breakpoint/jitbp_elf.test:3
+# RUN: %clang -g -S -emit-llvm -o %t.ll %p/Inputs/jitbp.cpp
+# RUN: %lldb -b -o 'b jitbp' -o 'run -jit-kind=mcjit %t.ll' lli | FileCheck %s
With these args, clang shouldn't optimize away `jitbp()`

  rG LLVM Github Monorepo



More information about the lldb-commits mailing list