[Lldb-commits] [PATCH] Fix failing tests caused by lack of default debug info

jingham at apple.com jingham at apple.com
Fri Feb 27 14:17:30 PST 2015


BTW, be careful if you start requiring various bits of system debug information to get debugging to work properly, since the stepping machinery uses "has debug information" and one of the criteria for whether "step in" stops in a function.  So when you add system debug info when running the testsuite, you may find errant failures in the testsuite.  You can usually work around this by setting "target.process.thread.step-avoid-libraries" to include the system libraries for which you've added debug information. You don't need to worry about this for the code in the "std" namespace, however, since we set "target.process.thread.step-avoid-regexp" to ^std by default, so we'll avoid stepping into std code in so far as we can.

Jim

> On Feb 27, 2015, at 1:46 PM, Reid Kleckner <rnk at google.com> wrote:
> 
> On Fri, Feb 27, 2015 at 12:51 PM, Siva Chandra <sivachandra at google.com> wrote:
> With reference to rnk's comment: "... but eventually we need to resolve this in LLDB"
> 
> All that I see in the DWARF generated by clang-3.5 is this:
> 
> < 2><0x00000037>      DW_TAG_class_type
> 
>   DW_AT_name                  "basic_string<char, std::char_traits<char>, std::allocator<char> >"
>   DW_AT_declaration           yes(1)
> 
> In which case, what can LLDB do?
> 
> Right, use the installed libstdc++-dbg symbols. But it is also reasonable to make sure LLDB tests don't fail if that package is installed, so this is a good patch to have. 
> _______________________________________________
> lldb-commits mailing list
> lldb-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/lldb-commits





More information about the lldb-commits mailing list