Repository navigation
x86 diff in displacement on data is ignored #422
Description
gMallocs is in .data, its a ->field_38 access is treated as without the member access.
Not sure when this regression started, the "Function Reloc Diffs" dropdown doesn't seem to do anything to help, maybe it's refactoring done some time ago is what broke it.
Extern diffs seem to work still
Assembled object
objdiffissue_displacement.zip
Assembly:
ml /c /coff /Zf /Cx /Cp objdiffissue_displacement.asm
.386P
.model flat
.stack 4096
assume fs:nothing
extern gExtMallocs:dword
.data
gMallocs DD ?
.code
_withdisplacement proc
mov esi, [gMallocs+38h]
ret
_withdisplacement endp
_nodisplacement proc
mov esi, [gMallocs+0h]
ret
_nodisplacement endp
_extwithdisplacement proc
mov esi, [gExtMallocs+38h]
ret
_extwithdisplacement endp
_extnodisplacement proc
mov esi, [gExtMallocs+0h]
ret
_extnodisplacement endp
end
Activity
tomsons26 commentedon Oct 18, 2026
v3.8.0 broke this, works in v3.7.3@LagoLunatic changed how reloc code works with this version, that's likely where this broke
Hmm no im seeing the issue in older versions too, did this ever work then...
LagoLunatic commentedon Oct 18, 2026
You only uploaded one test object, the target with the +0x38. Need both objects to test this.
LagoLunatic commentedon Oct 18, 2026
Or wait, are you saying that the one object you uploaded should be diffed with itself, like this?
If so I don't see any bug, on default settings it shows the expected diff. You must have messed with your config settings.
tomsons26 commentedon Oct 18, 2026
I do have
"options": {
"functionRelocDiffs": "data_value"
}
in the objdiff.json, seems it's overriding the dropdown then?
Thought it was a preference not a unconditional override
LagoLunatic commentedon Oct 18, 2026
I wasn't aware you could override user preferences with objdiff.json like that. Apparently it was added in #263, and mostly still works, but the visual indicator in the GUI that the option has been overriden does not seem to appear anymore for some reason. So I'm closing this, but I'll open an issue for the visual regression.