| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
thrown exception
Probably should have done this sooner, just lets the user continue analysis, rtti exceptions are continuable from the view of analysis.
|
|
Fixes https://github.com/Vector35/binaryninja-api/issues/6837
|
|
Found when a vtable is constrained to a size less than the base vft, indicating a bad base vft associating most likely.
Fixes https://github.com/Vector35/binaryninja-api/issues/6840
|
|
|
|
We don't do enough with the lifted il != non lifted il to justify the bound.
This makes modifying IL much less work as the historical lifted il bound is gone.
|
|
|
|
This is being done to reduce complexity in function signatures, specifically many of the strings we are passing ultimately should be new types themselves instead of "just strings", things such as type ids.
Another place which was confusing was dealing with filesystem related APIs, this commit turns most of those params into a stricter `Path` type.
This is bringing the rust api more inline with both python and C++, where the wrapper eagerly converts the string into the languages standard string type.
Special consideration must be made for symbols or other possible non utf-8 objects.
This commit will be followed up with one that adds the `IntoCStr` bound on API's we want to keep as invalid utf-8 so we can for example, get section by name on a section with invalid utf-8.
|
|
- Removed `to_string` shortcut from `BnString`.
- Misc formatting
|
|
|
|
conversion can be tolerated
Still need to go and audit all usage, but realistically the most important places to give the user control are with symbols, where the data can come from non utf8 sources
This is still incomplete, I just looked for usage of -> BnString so any other variant was omitted.
|
|
|
|
We need to do this for RTTI info that resides in DefaultSectionSemantic sections, which will happen for some binaries where the data is placed in non-generic sections.
|
|
GCC emits a leading `*` to indicate that the type info is internal and
its name can be compared via pointer equality. It is not part of the
type name. This was only being handled when followed with `N`, but it
can apply to any mangled name.
Additionally, this updates some `std::string::find(...) == 0` calls in
the adjacent code to use `std::string::rfind(..., 0) == 0` as that bails
out of the string comparison as soon as the prefix does not match,
rather than continuing to search the entire string.
|
|
32-bit ELF binaries that are dynamically linked to the C++ runtime may
use a copy relocation for the vtable. The vtable itself will be defined
in the `.bss` section, and the copy relocation will cause the dynamic
linker to populate it at load time from the C++ runtime library.
Detect this by looking for a symbol pointing to the start of the vtable data,
two pointers before the vtable address.
|
|
|
|
|
|
This is apart of https://github.com/Vector35/binaryninja-api/issues/6678
|
|
introduced analysis is processed before pipeline completion.
|
|
Finish verifying VMI base class info before applying data variable and symbol
|
|
|
|
|
|
|
|
- Do not add binary base to function address twice when a symbol with that function's raw name already exists
- Load eh_frame/debug_frame from base bv instead of debug bv and make calculated cie offset ranges relative to bv start
- Fix dwarf raw name resolution not resolving specification
- Try to load eh_frame/debug_frame from both raw and normal views in dwarf import
|
|
|
|
The vft will have a few fields above it, we did not verify that when we read those fields that they were readable.
Fixes https://github.com/Vector35/binaryninja-api/issues/6694
|
|
Fixes https://github.com/Vector35/binaryninja-api/issues/6692
|
|
|
|
|
|
|
|
|
|
|
|
Also adds more information to the progress text when processing vfts for both msvc and itanium
|
|
Some missing checks in case there is a billion segments (see .obj files)
|
|
IDK why this was not done prior, leaking the returned core activity and workflow object.
|
|
|
|
|
|
|
|
If the background task gets stuck, this is the commit to ponder at
|
|
Not even going to check the git blame i know this was committed at 3 in the morning
|
|
32bit support still needs some fixes for the VMI offsets and other things, but its better than it was before.
Likely need to do a little refactoring after this release to make the code less horrendous, but its fine for now.
|
|
Unlikely to be the last for initial analysis, still a good idea to check.
|
|
|
|
We really need a IsValidRange function
|
|
|
