[Lldb-commits] [PATCH] D73921: Assert that a subprogram should have a name when parsing DWARF

Mon Feb 10 15:28:04 PST 2020

jingham added a comment.

In D73921#1855961 <https://reviews.llvm.org/D73921#1855961>, @davide wrote:

> DWARFASTParserClang looks to me the wrong layer to fix this. Why can't this be caught in the generic DWARF Parser?
>  I also believe that it's better if dwarfdump -verify crashes on this, rather than lldb.

Many

In D73921#1868191 <https://reviews.llvm.org/D73921#1868191>, @JDevlieghere wrote:

> Given that this error is non-actionable, I don't see any value in diagnosing this LLDB. It is important to have this in dwarfdump, which does not detect this right now.
>
> It might be interesting to have LLDB run in a sort of "pedantic" mode which verifies all the DWARF it consumes with the dwarf verifier in LLVM. We have something similar in dsymutil which runs the verifier over the generated dSYM.

Note that many OS X developers never debug a dSYM build of their project.  They debug with .o files, then make a dSYM when they do their release builds.  And they probably don't look at the output of dsymutil amidst all the noise of a build.  So if we only do this in dsymutil, we are greatly narrowing the range of folks who might see & report this error to us.

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D73921/new/

https://reviews.llvm.org/D73921