* Move symbol name processing from diff step to read step
* Use min_by_key instead of sorted_unstable_by_key
* Sort all symbols by address when reading the object
This fixes an underreport bug in diff_bss_section.
* More symbol sorting logic on read
* Move is_name_compiler_generated into SymbolFlag & filter out File symbols
---------
Co-authored-by: Luke Street <luke@street.dev>
* Fix `__arraydtor$1234`s sometimes being swapped
The list of symbols is not necessarily sorted by address, so if there are multiple `__arraydtor$1234` functions it's possible that they're in reverse order on the right, which would cause them to be paired up incorrectly.
To fix this we sort by section index and symbol address after filtering down the list of symbols to ones that have matching names.
* Support pairing up symbols for anonymous classes
* Disallow exact-name pairups for compiler-generated symbols
* Implement diffing individual data symbols
* Remove unused code for diffing sections
* Data diff view: Make rows show offset within the symbol, not within the section
* Remove SelectedSymbol enum as it only has a single variant now
* Create fake data section symbols to allow diffing entire sections again
* Fix text sections not having their size zeroed out
* Update test snapshots
* Clean up code for inferring section symbol size
* Fix bug where PPC pool references weren't ignoring section symbols
* Update comment
* Always add unique section symbols for data sections
* Update test snapshots
* Remove unnecessary clone in format! call
* Auto-start mapping for unpaired data symbols
* Fix filtering out symbols from the target side that have a symbol with the same name and a `.NON_MATCHING` suffix.
- Target side: Show all the symbols except the `.NON_MATCHING` ones.
- Base side: Ignore all the `.NON_MATCHING` symbols and also ignore the ones with the same name without the suffix
* fmt
* comment
* tests
* fmt tests
* maybe this could fix wasm?
* Fix wasm?
* fmt
* Move `DiffSide` to `diff` mod
* Update the stuff the advisories CI told me to
* Improve automatic symbol pairing for nameless literals
* Fix data reloc diffing when the reloc points to an in-function static symbol
* Only pair up literals that match perfectly
* Clippy
* Do two separate passes when pairing up literals
* Fix partially-matching literal pairups not working right
* Remove duplicate $ splitting code
* Implement $ splitting for section names too
* Minor cleanup
Check raw code length when instructions have same disassembly
but different encodings. Marks as OpMismatch to indicate
encoding difference.
Fixes#242Fixes#233
* Fix data flow analysis for multiple text sections
* Data flow analysis results were only keyed by the symbol (function)
address. That doen't work if there are multiple text sections, the
result from the first function in one section will stomp the result
from the first function in another because both have address zero.
* Remove the ambiguity by keying off of the section address as well.
* Formatting
* Satisfy wasm build
* Clippy
* Formatting again
* Thought that section was the section address not the section number.
---------
Co-authored-by: Luke Street <luke.street@encounterpc.com>
LagoLunatic
Luke Street
Dávid Balatoni
Anghelo Carvajal
iFarbod
Luke Street
Toms
Mark Langen