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: Inconsistency between code and docs


On Thu, Oct 18, 2012 at 8:02 AM, Paulo Matos <pmatos@broadcom.com> wrote:
>
> I have found a strange inconsistency between code and docs for TARGET_ASM_NAMED_SECTION.
>
> Docs say (http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gccint/File-Framework.html#index-TARGET_005fASM_005fNAMED_005fSECTION-4472):
> " If decl is non-NULL, it is the VAR_DECL or FUNCTION_DECL with which this section is associated."
>
> However, in dwarf2out.c, line 21030 we have the following code:
>           tree comdat_key = get_identifier (ref->info);
>           /* Terminate the previous .debug_macinfo section.  */
>           dw2_asm_output_data (1, 0, "End compilation unit");
>           targetm.asm_out.named_section (DEBUG_MACRO_SECTION,
>                                          SECTION_DEBUG
>                                          | SECTION_LINKONCE,
>                                          comdat_key);
>
> This passes comdat_key to named_section hook which expected either a VAR_DECL or a FUNCTION_DECL.
> I noticed as I am using DECL_ATTRIBUTES inside the hook on decl and it failed with checking enabled as an IDENTIFIER_NODE (as passed by dwarf2out.c) is not a DECL.
>
> It seems to me to be a bug of dwarf2out.c since it should maybe generate a VAR_DECL instead of an IDENTIFIER_NODE. Is this the case?

I think the bug is in the documentation, and that
TARGET_ASM_NAMED_SECTION should accept an IDENTIFIER_NODE.

Ian


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