[Dwarf-discuss] DW_AT_frame_base

[Robinson, Paul]

A "location description [that] is a register operation" is the language in DWARF v3; in later versions, it is "a simple register location description." This means something like DW_OP_reg5, which is allowed in a location description but not in a DWARF expression.

Form DW_FORM_data4, value 0, would be interpreted as a location list reference. It is not by itself an address or a register location description.
--paulr

From: Dwarf-discuss <dwarf-discuss-bounces+paul.robinson=sony.com@lists.dwarfstd.org> On Behalf Of Vsevolod Alekseyev via Dwarf-discuss
Sent: Monday, September 18, 2023 10:08 AM
To: dwarf-discuss@lists.dwarfstd.org
Subject: [Dwarf-discuss] DW_AT_frame_base

Please help me interpret the DWARF spec regarding the DW_AT_frame_base. Quoted by spec v3, section 3.5.5, but similar wording in v4 and v5:

"A subroutine or entry point entry may also have a DW_AT_frame_base attribute, whose value is a location description that computes the "frame base" for the subroutine or entry point. If the location description is a register operation, the given register contains the frame base address. If the location description is a DWARF expression, the result of evaluating that expression is the frame base address. Finally, for a location list, this interpretation applies to each location expression contained in the list of location list entries."

So what does "location description that is a register operation" mean here? Since the option of a DWARF expression block is covered by the second option, that rather suggests to me that "register operation" is not a DWARF operation. I mean, the wording "if A is X, then Y. If A is P, then Q" usually means that P is distinct from X, right?

On a more practical note, I'm currently staring at a crash report with a DWARF attribute parsing failure. DWARF v3 Linux ELF binary, produced by NASM. DW_AT_frame_base, form DW_FORM_data4, value 0. The code assumes it's a loclist pointer, but the binary doesn't contain a loclist section.

Zero as a loclist pointer in a v3 binary could make sense. As a "register operation" - I'm not sure. Could be a compiler quirk, but I'm admitting the possibility that I'm misreading something.



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dwarfstd.org/pipermail/dwarf-discuss/attachments/20230918/fbae38c4/attachment.htm>