Sanitize ELF Sections API #27
Conversation
ELF Sections are kinda 1-indexed in that while there is a Section 0, it is always of type SHT_NULL and is otherwise invalid to reference. Further, later ELF extensions have re-used these unused bits for extensions, notably for ELF files with more than 0xff00 sections. This patch does the following: - Prevents section 0 from being returned by "Sections", since it isn't actually a real section and treating it as such would be incorrect if the section header stores extension data (before it just doesn't make a difference) - As a result indices into `Sections` now actually match the ELF section number rather than being off by one. - For good measure, implement the ELF large sections extension and add a test to make sure we can handle a file with as many sections.
| for i = 0:0xffff | ||
| print(f, ".section .text$i\nfoo$i:\nret\n") | ||
| end | ||
| end |
There was a problem hiding this comment.
Is this file actually used for anything?
There was a problem hiding this comment.
I think it's supposed to be used to generate largesectcount.S but it seems to be outputting to test.S instead.
Would be nice if it compiled it to largesectcount.o as well.
|
As an aside, if possible, I'd like it if we can avoid embedding the large test files into the git repository, as they'll actually significantly inflate user install sizes. We can either store only the generator |
|
Can you rebase on latest |
ELF Sections are kinda 1-indexed in that while there is a Section 0,
it is always of type SHT_NULL and is otherwise invalid to reference.
Further, later ELF extensions have re-used these unused bits for
extensions, notably for ELF files with more than 0xff00 sections.
This patch does the following:
actually a real section and treating it as such would be incorrect
if the section header stores extension data (before it just doesn't
make a difference)
Sectionsnow actually match the ELF sectionnumber rather than being off by one.
a test to make sure we can handle a file with as many sections.