lldb-test doesn’t actually exercise any of this. It’s a pretty new addition and doesn’t even support pdb yet as far as I know. So when you say it fails without the other changes, but passes with this, I think you must be talking about some test that runs via check-lldb, or some unittest.<br><br>To be clear, I’m talking about the executable lldb-test.exe, which was recently introduced as a way to run FileCheck tests. I was thinking you could add a flag to it and call it as<br><br>lldb-test.exe symbol -functions foo.exe<br><br>And have it just print function signatures and FileCheck the output for a function you’ve written that takes no arguments. That should crash without this patch but succeed with this patch alone right?<br><br>Since lldb-test is a tool whose sole purpose is printing stuff for testing purposes, the format isn’t too important, the important thing is just that it provides a really easy avenue for exercising this code.<br><br>LMK if this makes sense<br><div class="gmail_quote"><div dir="ltr">On Thu, Dec 21, 2017 at 1:53 PM Aaron Smith via Phabricator <<a href="mailto:reviews@reviews.llvm.org">reviews@reviews.llvm.org</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">asmith added a comment.<br>
<br>
This fix is part of a larger set of changes to retrieve the type for a function signature and I don't see how to test for this without those changes. With all the other changes, lldb-test fails without this fix and passes with it. So it's implicitly already tested.<br>
<br>
<br>
Repository:<br>
rL LLVM<br>
<br>
<a href="https://reviews.llvm.org/D41427" rel="noreferrer" target="_blank">https://reviews.llvm.org/D41427</a><br>
<br>
<br>
<br>
</blockquote></div>