clayborg added a comment.

I will make the fixes and also test out mapping "" to "." as suggested.



================
Comment at: source/Target/PathMappingList.cpp:194
+      // path and any path we appended would end up being relative.
+      fixed.SetFile(path_ref, false);
+    } else {
----------------
When I thought about it, I chose to not convert to "." for path remapping. I 
think people would expect if the remap "" to "/prefix" that "/prefix" would be 
prepended to each path no matter what it is. Of course users could just specify 
"/" if they wish for a prefix. I could see this going either way. Let me know 
what you think.


================
Comment at: unittests/Utility/PathMappingListTest.cpp:86
+    {"/old/foo.c/.", "/new/old/foo.c"},
+    {"/old/./foo.c", "/new/old/foo.c"},
+  };
----------------
labath wrote:
> How does this work for relative paths? I take it `foo.c` should be remapped 
> to `/new/foo.c` ? Can you add a test for that?
It doesn't work. If "foo.c" would map to "/new/foo.c", we can unmap it 
correctly since it would unmap to "/foo.c". Both "foo.c" and "/foo.c" would map 
to to "/new/foo.c" and then we can only unmap the latter correctly. 

This might bode well for saying that "" should map to "." actually. Then we 
won't run into this situation. I will test things out with "" mapping to "." 
and see how things go.


https://reviews.llvm.org/D47021



_______________________________________________
lldb-commits mailing list
lldb-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits

Reply via email to