[cfe-dev] Should "imports" list in DICompileUnit have duplication of DIImportedEntity?

[David Blaikie via cfe-dev]

Yeah, I can't see any reason we'd benefit from them either.

On Wed, Mar 2, 2016 at 11:37 AM, Adrian Prantl <aprantl at apple.com> wrote:

>
> > On Mar 2, 2016, at 12:59 AM, Aboud, Amjad <amjad.aboud at intel.com> wrote:
> >
> > Hi,
> > It looks like Clang FE does not treat imported entities list (“imports”)
> as a set that has no duplication.
> > In the below simple example, imported entities list contains a
> duplication for same DIImportedEntity, which will end up having duplication
> of this imported entity in the dwarf debug info section (see below).
>
> It looks like we get into this situation easily since
> DIBuilder::createImportedModule will add every namespace mentioned in a
> using directive.
>
> > Should "imports" list in DICompileUnit have duplication of
> DIImportedEntity?
> Right now I can’t think of a good reason where this would be useful. And
> if it is useful, we can always use distinct MDNodes to work around the
> problem.
> > Do you think we should fix Clang to assure the "imports" list is a
> unique set?
>
> I would also take a look at the DIBuilder interface — it may or may not be
> better to unique them there.
> >
> > Notice that there is no added value to have the same entity duplicated!
> >
>
> thanks for noticing!
>
> adrian
>
> > Thanks,
> > Amjad
> >
> >
> > The example (using the attached files):
> >
> > Command-line: clang -cc1 -S -emit-llvm -o - -debug-info-kind=limited
> TestIM.cpp
> > !0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !1,
> producer: "clang version 3.9.0 (trunk 261634) (llvm/trunk 261633)",
> isOptimized: false, runtimeVersion: 0, emissionKind: 1, enums: !2, imports:
> !3)
> > !1 = !DIFile(filename: " <stdin>", directory: "\\")
> > !2 = !{}
> > !3 = !{!4, !4}  <----------------------------  Bug: same entity appears
> twice in the imported entities list!
> > !4 = !DIImportedEntity(tag: DW_TAG_imported_declaration, scope: !5,
> entity: !7, line: 8)
> > !5 = !DINamespace(name: "Y", scope: null, file: !6, line: 7)
> >
> >
> > Command-line: clang -cc1 -S -emit-obj -o - -debug-info-kind=limited
> TestIM.cpp | llvm-dwarfdump.exe -debug-dump=info –
> > 0x0000000b: DW_TAG_compile_unit [1] *
> >               DW_AT_producer [DW_FORM_strp]     ( .debug_str[0x00000000]
> = "clang version 3.9.0 (trunk 259349) (llvm/trunk 259357)")
> >               DW_AT_language [DW_FORM_data2]    (DW_LANG_C_plus_plus)
> >               DW_AT_name [DW_FORM_strp] ( .debug_str[0x00000037] = "
> <stdin>")
> >               DW_AT_stmt_list [DW_FORM_sec_offset]      (0x00000000)
> >               DW_AT_comp_dir [DW_FORM_strp]     ( .debug_str[0x0000004e]
> = " \\ ")
> >
> > 0x0000001e:   DW_TAG_namespace [2] *
> >                 DW_AT_name [DW_FORM_strp]       ( .debug_str[0x00000074]
> = "Y")
> >                 DW_AT_decl_file [DW_FORM_data1] ("TestIM.h")
> >                 DW_AT_decl_line [DW_FORM_data1] (7)
> >
> > 0x00000025:     DW_TAG_imported_declaration [3]
> >                   DW_AT_decl_file [DW_FORM_data1]       ("TestIM.h")
> >                   DW_AT_decl_line [DW_FORM_data1]       (8)
> >                   DW_AT_import [DW_FORM_ref4]   (cu + 0x003b =>
> {0x0000003b})
> >
> > 0x0000002c:     DW_TAG_imported_declaration [3]
>  <----------------------------  duplication of 0x00000025
> >                   DW_AT_decl_file [DW_FORM_data1]       ("TestIM.h")
> >                   DW_AT_decl_line [DW_FORM_data1]       (8)
> >                   DW_AT_import [DW_FORM_ref4]   (cu + 0x003b =>
> {0x0000003b})
> >
> > 0x00000033:     NULL
> >
> > ---------------------------------------------------------------------
> > Intel Israel (74) Limited
> >
> > This e-mail and any attachments may contain confidential material for
> > the sole use of the intended recipient(s). Any review or distribution
> > by others is strictly prohibited. If you are not the intended
> > recipient, please contact the sender and delete all copies.
> >
> > <TestIM.H><TestIM.CPP>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20160302/0678be98/attachment.html>