> On Jan 20, 2015, at 11:07 AM, David Blaikie <[email protected]> wrote: > > My vague recollection from the previous design discussions was that these > module references would be their own 'unit' COMDAT'd so that we don't end up > with the duplication of every module reference in every unit linked together > when linking debug info? > > I think in my brain I'd been picturing this module reference as being an > extended fission reference (fission skeleton CU + extra fields for users who > want to load the Clang AST module directly and skip the split CU).
Apologies for letting this rest for so long. Your memory was of course correct and I didn’t follow up on this because I had convinced myself that the fission reference would be completely sufficient. Now that I’ve been thinking some more about it, I don’t think that it is sufficient in the LTO case. Here is the example from the http://lists.cs.uiuc.edu/pipermail/cfe-dev/2014-November/040076.html: foo.o: .debug_info.dwo DW_TAG_compile_unit // For DWARF consumers DW_AT_dwo_name ("/path/to/module-cache/MyModule.pcm") DW_AT_dwo_id ([unique AST signature]) .debug_info DW_TAG_compile_unit DW_TAG_variable DW_AT_name "x" DW_AT_type (DW_FORM_ref_sig8) ([hash for MyStruct]) In this example it is clear that foo.o imported MyModule because its DWO skeleton is there in the same object file. But if we deal with the result of an LTO compilation we will end up with many compile units in the same .debug_info section, plus a bunch of skeleton compile units for _all_ imported modules in the entire project. We thus loose the ability to determine which of the compile units imported which module. I think it really is necessary to put the info about the module imported into the compile unit that imported it. Or is there a way to do this using the fission capabilities that I’m not aware of? -- adrian > > [rambling a bit more along those lines: > This would work fine in the case of the module (now an object file) > containing all the static debug info > The future step, when we put IR/object code in a module to be linked into the > final binary, we could put the skeleton CU in that object file that's being > linked in (then we wouldn't need to COMDAT it) or, optionally, link in the > debug info itself (skipping the indirection through the external file) if a > standalone debug info executable was desired] > > On Tue, Jan 20, 2015 at 9:39 AM, Adrian Prantl <[email protected]> wrote: > As a complementary part of the module debugging story, here is a proposal to > list the imported modules in the debug info. This patch is not about > efficiency, but rather enables a cool debugging feature: > > Record the clang modules imported by the current compile unit in the debug > info. This allows a module-aware debugger (such as LLDB) to @import all > modules visible in the current context before evaluating an expression, thus > making available all declarations in the current context (that originate from > a module) and not just the ones that were actually used by the program. > > This implementation uses existing DWARF mechanisms as much as possible by > emitting a DW_TAG_imported_module that references a DW_TAG_module, which > contains the information necessary for the debugger to rebuild the module. > This is similar to how C++ using declarations are encoded in DWARF, with the > difference that we're importing a module instead of a namespace. > The information stored for a module includes the umbrella directory, any > config macros passed in via the command line that affect the module, and the > filename of the raw .pcm file. Why include all these parameters when we have > the .pcm file? Apart from module chache volatility, there is no guarantee > that the debugger was linked against the same version of clang that generated > the .pcm, so it may need to regenerate the module while importing it. > > Let me know what you think! > -- adrian > > > _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
