radare2/test
2023-03-03 10:14:51 +01:00
..
bench Add quickjs benchmark script 2022-01-06 22:54:55 +01:00
db Add CLLf command as an alias for 'list' and fix @@i ##dwarf 2023-03-03 00:03:38 +01:00
fuzz Clean-up ASN.1 parser and update x509 fuzzing 2023-01-25 01:44:05 +01:00
notworking_db
ravc2_git_branch_test@a3ea888e31 Fix some bugs in projects ##projects 2022-12-06 19:08:23 +01:00
scripts Fix webtest script 2022-05-08 10:44:53 +02:00
unit Remove RList.get_top and RList.get_bottom 2023-03-03 10:14:51 +01:00
.gitignore Add option to log r2r test results and publish it in the CI (#17964) ##test 2020-11-24 03:43:41 +01:00
Makefile Update sdb to remove double include paths ##build 2022-11-19 01:05:18 +01:00
README.md Fix typos in different locations 2023-01-16 18:41:04 +01:00
run_unit.sh

Radare2 Regression Test Suite

A set of regression tests for Radare2 (http://radare.org).

Originally based on work by and now in collaboration with pancake.

Directory Hierarchy

  • db/: The tests sources
  • unit/: Unit tests (written in C, using minunit).
  • fuzz/: Fuzzing helper scripts
  • bins/: Sample binaries (fetched from the external repository)

Requirements

  • Radare2 installed (and in $PATH or set the R2 environment).
  • r2pipe tests require Python and r2pipe (in CI uses python3 and r2pipe from git, but users may be good with latests releases)
  • Valgrind (optional).

Usage

  • To run all tests, use make -k all.
  • To execute only the unit tests use make -k unit_tests.

Failure Levels

A test can have one of the following results:

  • success: The test passed, and that was expected.
  • fixed: The test passed, but failure was expected.
  • broken: Failure was expected, and happened.
  • failed: The test failed unexpectedly. This is a regression.

Reporting Radare2 Bugs

Please do not post Radare2 bugs on the r2-regressions github tracker. Instead use the official r2 tracker:

https://github.com/radareorg/radare2/issues?state=open

Writing Assembly tests

Example tests for db/asm/*:

General format:
type "assembly" opcode [offset]

	type:
		* a stands for assemble
		* d stands for disassemble
		* B stands for broken
		* E stands for cfg.bigendian=true

	offset:
		Some architectures are going to assemble an instruction differently depending
		on the offset it's written to. Optional.

Examples:
a "ret" c3
d "ret" c3
a "nop" 90 # Assembly is correct
dB "nopppp" 90 # Disassembly test is broken

You can merge lines:

adB "nop" 90

acts the same as

aB "nop" 90
dB "nop" 90

    The filename is very important. It is used to tell radare which architecture to use.

    Format:
    arch[[_cpu]_bits]

Example:
x86_32 means -a x86 -b 32
    arm_v7_64 means what it means

Writing JSON tests

The JSON tests db/json are executed on 3 standard files (1 ELF, 1 MachO, 1 PE). The tests need to be working on the 3 files to pass.

Commands tests


Example commands tests for the other db/ folders:

NAME=test_db
FILE=bins/elf/ls
CMDS=<<EXPECT
pd 4
EXPECT=<<RUN
        ;-- main:
        ;-- entry0:
        ;-- func.100001174:
        0x100001174      55             Push rbp
        0x100001175      4889e5         Mov  rbp, rsp
        0x100001178      4157           Push r15
RUN
  • NAME is the name of the test, it must be unique
  • FILE is the path of the file used for the test
  • ARGS (optional) are the command line argument passed to r2 (e.g -b 16)
  • CMDS are the commands to be executed by the test
  • EXPECT is the expected output of the test
  • BROKEN (optional) is 1 if the tests is expected to be fail, 0 otherwise
  • TIMEOUT (optional) is the number of seconds to wait before considering the test timeout

You must end the test by adding RUN keyword

Advices

  • For portability reasons Do not use shell pipes, use ~
  • dont use pd if not necessary, use pi
  • All tests use the UTC timezone for consistency.

License

The test files are licensed under GPL 3 (or later).