Adjust symtab section header #31
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a145bf1f50
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
katsyoshi
force-pushed
the
adjust-symtab-section-header
branch
from
2cd8a69 to
3a54b40
Compare
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3a54b40262
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| name, info, link, entsize = (name_offset.nil? ? 0 : name_offset), header.info, header.link, 0 | ||
| if header.type == :symtab | ||
| info = section.body.size | ||
| link = strtabndx |
There was a problem hiding this comment.
In ELF, a symtab section’s sh_info must be the index of the first non-local symbol (i.e., the count of local symbols). Here info = section.body.size sets it to the total symbol count, so any global symbols are effectively marked as local to consumers that rely on sh_info. This shows up whenever the input has globals (e.g., the main symbol from the assembler or GCC-produced objects), and it can cause tools/linkers to misclassify or ignore globals.
Useful? React with 👍 / 👎.
| write_order << @elf_obj.find_by_name(".text") | ||
| write_order << @elf_obj.find_by_name(".symtab") | ||
| write_order << @elf_obj.find_by_name(".strtab") | ||
| write_order.concat(@elf_obj.select_by_names([/\.rel\./, /\.rela\./])) | ||
| write_order << @elf_obj.find_by_name(".shstrtab") |
There was a problem hiding this comment.
The write order now includes every .rel.*/.rela.* section but only keeps .text, .symtab, .strtab, and .shstrtab. For objects that also have relocations for other sections (e.g., GCC adds .rela.eh_frame alongside .eh_frame), this emits relocation sections whose referenced target section is omitted from the output. That leaves sh_info pointing at a section index that doesn’t exist in the new section table, producing an invalid ELF that can confuse tooling. Either include the referenced sections or drop relocations whose targets aren’t being written.
Useful? React with 👍 / 👎.
Summary
Testing