This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [LTO] LTO breaks if debug info is stripped from object files


On Nov 18, 2007 12:11 PM, Mark Mitchell <mark@codesourcery.com> wrote:

>  In particular, in the
> current implementation, none of the debug information for the bodies of
> functions (i.e., that describing local variables, line numbers for
> statements, etc.) is required for LTO processing.  The LTO use is
> confined to global variables, functions, types, etc.

This is an issue that I'm looking at.  I raised it with Nathan, but haven't
heard back from him.  It appears that, at present, generating LTO info
forces generation of all of the debug information  that would be produced
with '-g', including line number tables and frame tables.  In fact, debug
debug info generation is turned on by -flto in the options processing code.
There are some adjustments due to LTO, including emitting some of the
debug info earlier, as well as a few others that I suspect result in broken
debug info.  I've been looking at disentangling the code paths, that is,
viewing the writing of LTO info as distinct from writing debug info, though
it needs to be coordinated in order to avoid duplication.

--Bill


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]