A bunch of issues to discuss/move/reorder

[radare]

R2 TODO

• Allow to seek to branch N like in visual, but from cmdline
• refactor vmenus.c -> refresh function must be redefined for each menu
• if no debugger supported, say so... r_io must ?

BUGS

• opening a file from inside r2 doesnt clears internal data (strings..)

• allow to hook r_asm_disassemble and assemble with custom callbacks

  • extend a disassembler with own instructions.

• register renaming (per-instruction or ranges)

  • r_parser fun? a specific asm.parser plugin that does all this tricks?

• Display getsym() stuff in rabin2, not only legit syms

• dmi command must read from memory if no file path provided

  • rabin from memory ftw, to get libnames of dll, so..

• add support for sign/unsigned registers..or at least a way to cast them

• use r_anal_value everywhere

• diff code analysis

  • diff two programs 1st level:
    • check all functions EQUAL, DIFFERENT, REMOVED, ADDED
    • check all symbols
    • check all imports
    • check all strings 2nd level:
    • basic block level diffing (output in graph mode)

Random

• Implement rap:// upload/download protocol commands (maybe just system() with rsc2+wget?

• Reimplement or fix the delta diffing in C - first we need to do it for ired..

• Ranged/scrollable zoom mode

• Add support for STATIC_PLUGINS in r_lang

  • r_lang_define is implemented in lang.c, but requires the collaboration of the plugins to properly setup the environment for the script execution.
  • Add support for STATIC_PLUGINS in r_lang
  • dlerror(/usr/lib/radare2/lang_perl.so): libperl.so: cannot open shared object file: No such file or directory This issue is fixed by setting LD_LIBRARY_PATH...looks like dlopen ignores rpath

• gdiff

  • graph based fingerprints? (cyclomatic complexity...)

• rcore

  • r_asm can reduce cpu without disasm on fixed size ops archs.

Debugger

• Step until end of frame (stack pointer restored) (store sp, check if nsp>sp)
• stepover waits for one unknown event that cannot be stopped
• code injection facilities? (wtf? insert, execute, restore)
• Trace contents of buffers: filter search results..? cc 8080 @@ hit* .. check for values that has changed.
• Record trace of register status for each function when running
  • r_reg_arena_copy();
• All threads must be stopped when a breakpoint is handled..
• Floating point registers
• MMX/XMM/DRX control

pancake

• Implement PTRACE_BLOCK on Linux
• fork/clone child . inject code to create new threads or pids
• Functions in r_util to get lil/big ut8,16,32 from ut8*
  • already done..must find better names probably
• Implement RAnalCall (analyze function arguments, return values, propagate types..)
  • define number of arguments for given function
  • warn if signature and analysis differs in number of args or so..
  • when calling a function
  • identify arguments passed and compare with arguments required
  • if they do not match: we need to warn/ask user/store multiple options
    • function signature comparsion if they dont match r_anal_fcn_cmp (anal, f1, f2);

Refactoring

• Is RCore->block and blocksize a RBuf ? refactor!11
• Discuss missing r_core_sysenv_update in core/file.c:33
• Add RLog API.. pipeable to disk and stderr..also hookable ..cool for ui (partially done)
• Redesign core/disasm.c to provide a pluggable api
• Move 'r_syscall_t' stuff into r_debug (sync r_core)
• Implement r_bind api to link multiple pointers core->asm = r_bind_set (core->asm->bind, r_asm_new ());

To wipe

• imho we should not implement this:
  • Implement BLOCK in r_core_sysenv_begin|end ()

Optimizations

• Performance
  • cons_visual_write() should do a single write instead of one per line
  • Refactor get_sym() and so on...
  • TODO: make elf/pe get_os() and others return const and not strdup
  • RAnalValue must be static, not ref
  • save memory and accelerate code analysis
  • basicblock signatures must be just pointers to a big buf

Future

• memset0 the op before calling the plugin analysis -- not really that is a performance cost..
• Implement r_flag_unset_i () ftw
• search.kwidx must be search.lastidx or search.idx ?
• rax2 -k by default?
• r_cons_visual_write_tail() -> fill end of screen with spaces \o/
• Add support for 'expect' like foo in rarun2
  • make rarun live in a lib.. or at least be usable from r2
• use centralized pubsub or memcached to sync data // redis?
• r_file_slurp should work fine for big files (not prio) r_file_slurp_buf?
  • mmap if supported - add r_file_mmap ? - read file in blocks instead of the whole file in a single syscall
• Realign flags when using project in debug mode
• FileDescriptors: dd -- copy from !fd in r1
• metaflags? support to define relations between flags (flag hirearchies) r_flagtree
  • r_flags should have a tree construction to access to them faster
    • btree? following pointers like bigger,smaller { struct r_flag_t bigger, smaller; }
    • hooks r_flag_add to recalculate in r_flag_optimize(), bigger/smaller pointers
    • hooks r_flag_del to recalculate too.
  • the r_flag_get by string should have another construction with btree for the string of the name

Syscall

• Same for IO ports
  • 0x378 : lp1 example: outb(lp1,33); lp1<=33 lp1=>eax /dev/port: lseek 33, write 20

GRUB

• Reduce the number of needed include files
• Add a sync method
• Use libtrum (from netbsd?)

[XVilka]

Long term, not for this release I think.

[Maijin]

cleaned some already done

[Maijin]

Well at this stage, probably everyone need to check out especially @radare , not just me ^_^

[radare]

Had to try :p

On 24 Sep 2018, at 10:06, Maijin notifications@github.com wrote:

Well at this stage, probably everyone need to check out especially @radare , not just me ^_^

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[Maijin]

I did some cleaning already, but some of the rest is mostly decisions

[radare]

@XVilka ^