fix(regsort): drop self-assignments through register aliases (#2795)

* fix(regsort): drop self-assignments through register aliases

The self-assignment drop compared register names only (k != v), so an
assignment through an alias (e.g. {'ax': 'eax'} with registers.native32)
was not recognized as a no-op.  It then surfaced as a spurious
1-register dependency cycle, which either raised PwnlibException with
xchg=False or emitted a useless 'mov' of a register into its own alias.

With xchg=False and other registers present, the bogus cycle also
forced the temporary-register code path, reordering unrelated moves and
silently leaving registers with wrong values (see the linked issue).

Compare the mapped names instead, like the rest of the function.

Fixes #2794

Co-authored-by: rahul05ranjan <[email protected]>

* docs(changelog): add regsort register-alias self-assignment fix entry

---------

Co-authored-by: Rahul Ranjan <[email protected]>

.github/workflows/ci.yml

@@ -22,8 +22,9 @@ jobs:
           - 3001:3001  # https://libc.rip/ proxy cache
           - 3002:3002  # http://archive.ubuntu.com/ proxy cache
           - 3003:3003  # https://gitlab.com/ proxy cache
+          - 3004:3004  # https://debuginfod.ubuntu.com/ proxy cache
     env:
-      DEBUGINFOD_URLS: http://localhost:3000/
+      DEBUGINFOD_URLS: "http://localhost:3000/ http://localhost:3004/"
       PWN_LIBCRIP_URL: http://localhost:3001/
       PWN_UBUNTU_ARCHIVE_URL: http://localhost:3002/
       PWN_GITLAB_LIBCDB_URL:  http://localhost:3003/
@@ -120,6 +121,7 @@ jobs:
           binutils-aarch64-linux-gnu \
           binutils-arm-linux-gnueabihf \
           binutils-mips-linux-gnu \
+          binutils-mipsel-linux-gnu \
           binutils-msp430 \
           binutils-powerpc-linux-gnu \
           binutils-s390x-linux-gnu \
@@ -148,6 +150,8 @@ jobs:
         pip install --upgrade wheel build
         pip install --upgrade flake8 appdirs
         pip install --upgrade --editable .
+        # Pin to known working version until 2.1.5 is out. https://github.com/unicorn-engine/unicorn/issues/2134
+        pip install 'unicorn==2.1.2'
 
     - name: Sanity checks
       run:  PWNLIB_NOTERM=1 python -bb -c 'from pwn import *; print(pwnlib.term.term_mode)'
@@ -283,7 +287,7 @@ jobs:
       run: |
         pip install --upgrade pip
         pip install --upgrade --editable .
-    
+
     - name: Sanity checks
       run: |
         python -bb -c 'from pwn import *'

.github/workflows/pylint.yml

@@ -24,7 +24,7 @@ jobs:
     - name: PyLint
       run: |
         set -x
-        pip install pylint
+        pip install 'pylint<4'
         pip install --upgrade -e .
         run_pylint() { pylint --exit-zero --errors-only pwnlib -f parseable | cut -d ' ' -f2- | sed 's/line [0-9]\+/line XXXX/g'; }
         run_pylint > current.txt

.readthedocs.yaml

@@ -5,7 +5,7 @@ version: 2
 build:
   os: ubuntu-22.04
   tools:
-    python: "3"
+    python: "3.12"
 
 sphinx:
   configuration: docs/source/conf.py

CHANGELOG.md

@@ -10,8 +10,9 @@ The table below shows which release corresponds to each branch, and what date th
 | Version          | Branch   | Release Date           |
 | ---------------- | -------- | ---------------------- |
 | [5.0.0](#500-dev)  | `dev`    |
-| [4.15.0](#4150-beta)  | `beta`   |
-| [4.14.0](#4140-stable)  | `stable` | Jan 15, 2025
+| [4.15.0](#4150-stable)  | `stable` | Oct 12, 2025
+| [4.14.1](#4141)  |          | Mar 24, 2025
+| [4.14.0](#4140)  |          | Jan 15, 2025
 | [4.13.1](#4131)  |          | Sep 29, 2024
 | [4.13.0](#4130)  |          | Aug 12, 2024
 | [4.12.0](#4120)  |          | Feb 22, 2024
@@ -74,8 +75,80 @@ The table below shows which release corresponds to each branch, and what date th
 
 ## 5.0.0 (`dev`)
 
+- [#2598][2598] aarch64: Fix ABI definition
+- [#2419][2419] riscv: avoid compressed instructions (if you need compressed, use .option rvc)
+- [#2551][2551] Detect when kitty is being used as terminal
+- [#2519][2519] Drop Python 2.7 support / Require Python 3.10
+- [#2507][2507] Add `+LINUX` and `+WINDOWS` doctest options and start proper testing on Windows
+- [#2522][2522] Support starting a kitty debugging window with the 'kitten' command
+- [#2524][2524] Raise EOFError during `process.recv` when stdout closes on Windows
+- [#2526][2526] Properly make use of extra arguments in `packing` utilities. `sign` parameter requires keyword syntax to specify it.
+- [#2517][2517] Allow to passthru kwargs on `ssh.__getattr__` convenience function to fix SSH motd problems
+- [#2530][2530] Do NOT error when passing directory arguments in `checksec` commandline tool.
+- [#2529][2529] Add LoongArch64 support
+- [#2506][2506] ROP: fix `ROP(ELF(exe)).leave` is `None` in some ELF
+- [#2504][2504] doc: add example case for `tuple` (host, port pair) in `gdb.attach`
+- [#2546][2546] ssh: Allow passing disabled_algorithms keyword argument from ssh to paramiko
+- [#2538][2538] Add `ssh -L` / `ssh.connect_remote()` workaround when `AllowTcpForwarding` is disabled
+- [#2574][2574] Allow creating an ELF from in-memory bytes
+- [#2575][2575] Detect when Terminator is being used as terminal
+- [#2578][2578] Add gnome-terminal, Alacritty, Ttilix for run_in_new_terminal
+- [#2590][2590] Add support for finding corefiles under WSL2
+- [#2496][2496] Add linux ko file search support
+- [#2542][2542] Decode `_IO_*` flags in `FileStructure` member
+- [#2592][2592] pwnlib.config: Fix customization of `context.timeout`
+- [#2608][2608] Abort on `libcdb file libc.so --unstrip` if eu-unstrip is not installed
+- [#2611][2611] Cleanup `pwnlib.lexer` exports and imports
+- [#2610][2610] Fix `log.progress` ignoring `context.log_console`
+- [#2615][2615] tube/process: Fix redirecting stderr to stdout on Windows
+
+[2598]: https://github.com/Gallopsled/pwntools/pull/2598
+[2419]: https://github.com/Gallopsled/pwntools/pull/2419
+[2551]: https://github.com/Gallopsled/pwntools/pull/2551
+[2519]: https://github.com/Gallopsled/pwntools/pull/2519
+[2507]: https://github.com/Gallopsled/pwntools/pull/2507
+[2522]: https://github.com/Gallopsled/pwntools/pull/2522
+[2524]: https://github.com/Gallopsled/pwntools/pull/2524
+[2526]: https://github.com/Gallopsled/pwntools/pull/2526
+[2517]: https://github.com/Gallopsled/pwntools/pull/2517
+[2530]: https://github.com/Gallopsled/pwntools/pull/2530
+[2529]: https://github.com/Gallopsled/pwntools/pull/2529
+[2506]: https://github.com/Gallopsled/pwntools/pull/2506
+[2504]: https://github.com/Gallopsled/pwntools/pull/2504
+[2546]: https://github.com/Gallopsled/pwntools/pull/2546
+[2538]: https://github.com/Gallopsled/pwntools/pull/2538
+[2574]: https://github.com/Gallopsled/pwntools/pull/2574
+[2575]: https://github.com/Gallopsled/pwntools/pull/2575
+[2578]: https://github.com/Gallopsled/pwntools/pull/2578
+[2590]: https://github.com/Gallopsled/pwntools/pull/2590
+[2496]: https://github.com/Gallopsled/pwntools/pull/2496
+[2542]: https://github.com/Gallopsled/pwntools/pull/2542
+[2592]: https://github.com/Gallopsled/pwntools/pull/2592
+[2608]: https://github.com/Gallopsled/pwntools/pull/2608
+[2611]: https://github.com/Gallopsled/pwntools/pull/2611
+[2610]: https://github.com/Gallopsled/pwntools/pull/2610
+[2615]: https://github.com/Gallopsled/pwntools/pull/2615
+
+## 4.15.1
+
+- [#2694][2694] fix: pad bytes fields to correct field size in FileStructure
+- [#2701][2701] Fix `adb._build_date()` crash on devices with non-standard locale date strings
+- [#2721][2721] Allow running with Unicorn 2.1.[34] but throw when emulating MIPS
+- [#2773][2773] Disable debuginfod doctests in libcdb.unstrip_libc
+- [#2772][2772] fix(packing): `dd()` no longer crashes on a `bytearray` destination
+- [#2787][2787] gdb.corefile: Run gdb directly to prevent detach race
+- [#2795][2795] regsort: Fix self-assignment detection through register aliases
+
+[2694]: https://github.com/Gallopsled/pwntools/pull/2694
+[2701]: https://github.com/Gallopsled/pwntools/pull/2701
+[2721]: https://github.com/Gallopsled/pwntools/pull/2721
+[2773]: https://github.com/Gallopsled/pwntools/pull/2773
+[2772]: https://github.com/Gallopsled/pwntools/pull/2772
+[2787]: https://github.com/Gallopsled/pwntools/pull/2787
+[2795]: https://github.com/Gallopsled/pwntools/pull/2795
+
+## 4.15.0 (`stable`)
 
-## 4.15.0 (`beta`)
 - [#2508][2508] Ignore a warning when compiling with asm on nix
 - [#2471][2471] Properly close spawned kitty window
 - [#2358][2358] Cache output of `asm()`
@@ -93,6 +166,19 @@ The table below shows which release corresponds to each branch, and what date th
 - [#2502][2502] Fix loading ELF files without valid .dynamic section
 - [#2476][2476] Deprecate 'keepends' argument in favor of 'drop' in `tube.recvline*`
 - [#2364][2364] Deprecate direct commandline scripts invocation and exclude nonsense ones
+- [#2570][2570] Fix `pwn template` to skip Docker library extraction for statically linked binaries
+- [#2545][2545] SSH: fix download/upload with -1 exit status
+- [#2567][2567] Fix mistakenly parsing of ld-linux error messages.
+- [#2576][2576] regsort: respect register aliases
+- [#2595][2595] libcdb: Add Ubuntu and Debian debuginfod servers to default list
+- [#2593][2593] Use unicorn on macOS w/ SIP enabled
+- [#2587][2587] Support longer function names in Windows `getexport` shellcode
+- [#2596][2596] Ignore `colored_traceback` error when TERM envvar is unset
+- [#2579][2579] Fix poll error in `process.libs()` and clean up maps parsing
+- [#2602][2602] Allow setting debugger path via context.gdb_binary
+- [#2609][2609] Fix log level of child remotes of `server` tube
+- [#2612][2612] Fix lookup of binutils for `mipsel` architecture
+- [#2624][2624] Fix regression: gdbserver can't handle command-line argument containing whitespace
 
 [2508]: https://github.com/Gallopsled/pwntools/pull/2508
 [2471]: https://github.com/Gallopsled/pwntools/pull/2471
@@ -111,8 +197,39 @@ The table below shows which release corresponds to each branch, and what date th
 [2502]: https://github.com/Gallopsled/pwntools/pull/2502
 [2476]: https://github.com/Gallopsled/pwntools/pull/2476
 [2364]: https://github.com/Gallopsled/pwntools/pull/2364
+[2570]: https://github.com/Gallopsled/pwntools/pull/2570
+[2545]: https://github.com/Gallopsled/pwntools/pull/2545
+[2567]: https://github.com/Gallopsled/pwntools/pull/2567
+[2576]: https://github.com/Gallopsled/pwntools/pull/2576
+[2595]: https://github.com/Gallopsled/pwntools/pull/2595
+[2593]: https://github.com/Gallopsled/pwntools/pull/2593
+[2587]: https://github.com/Gallopsled/pwntools/pull/2587
+[2596]: https://github.com/Gallopsled/pwntools/pull/2596
+[2579]: https://github.com/Gallopsled/pwntools/pull/2579
+[2602]: https://github.com/Gallopsled/pwntools/pull/2602
+[2609]: https://github.com/Gallopsled/pwntools/pull/2609
+[2612]: https://github.com/Gallopsled/pwntools/pull/2612
+[2624]: https://github.com/Gallopsled/pwntools/pull/2624
 
-## 4.14.0 (`stable`)
+## 4.14.1
+
+- [#2451][2451] Show symbols defined to value 0 (start of file)
+- [#2533][2533] Fix installation on Python 3.5 and lower
+- [#2518][2518] fix: update apport coredump path handling for CorefileFinder
+- [#2552][2552] Fix memcpy shellcraft template off-by-one
+- [#2559][2559] Fix parsing corefile with missing auxv
+- [#2562][2562] Fix syntax highlighting of multiline comments in ASM
+- [#2565][2565] Exclude broken Unicorn
+
+[2451]: https://github.com/Gallopsled/pwntools/pull/2451
+[2533]: https://github.com/Gallopsled/pwntools/pull/2533
+[2518]: https://github.com/Gallopsled/pwntools/pull/2518
+[2552]: https://github.com/Gallopsled/pwntools/pull/2552
+[2559]: https://github.com/Gallopsled/pwntools/pull/2559
+[2562]: https://github.com/Gallopsled/pwntools/pull/2562
+[2565]: https://github.com/Gallopsled/pwntools/pull/2565
+
+## 4.14.0
 
 - [#2356][2356] Add local libc database provider for libcdb
 - [#2360][2360] Add offline parameter for `search_by_hash` series function
@@ -183,7 +300,7 @@ The table below shows which release corresponds to each branch, and what date th
 - [#2308][2308] Fix WinExec shellcraft to make sure it's 16 byte aligned
 - [#2279][2279] Make `pwn template` always set context.binary
 - [#2310][2310] Add support to start a process on Windows
-- [#2335][2335] Add lookup optimizations in DynELF 
+- [#2335][2335] Add lookup optimizations in DynELF
 - [#2334][2334] Speed up disasm commandline tool with colored output
 - [#2328][2328] Lookup using $PATHEXT file extensions in `which` on Windows
 - [#2189][2189] Explicitly define p64/u64 functions for IDE support
@@ -270,7 +387,7 @@ The table below shows which release corresponds to each branch, and what date th
 
 ## 4.11.0
 
-- [#2185][2185] make fmtstr module able to create payload without $ notation 
+- [#2185][2185] make fmtstr module able to create payload without $ notation
 - [#2103][2103] Add search for libc binary by leaked function addresses `libcdb.search_by_symbol_offsets()`
 - [#2177][2177] Support for RISC-V 64-bit architecture
 - [#2186][2186] Enhance `ELF.nx` and `ELF.execstack`
@@ -545,7 +662,7 @@ In memoriam — [Zach Riggle][zach] — long time contributor and maintainer of
 [fecf9f]: http://github.com/Gallopsled/pwntools/commit/fecf9f
 [1454]: https://github.com/Gallopsled/pwntools/pull/1454
 
-## 4.1.7 (`stable`)
+## 4.1.7
 
 - [#1615][1615] Fix aarch64 pushstr and pushstr_array
 

docs/requirements.txt

@@ -19,6 +19,6 @@ psutil
 requests>=2.5.1
 ropgadget>=5.3
 sphinx==1.8.6; python_version<'3'
-sphinx>=7.0.0; python_version>='3'
+sphinx>=8.1.3, <9; python_version>='3'
 sphinx_rtd_theme
 sphinxcontrib-autoprogram<=0.1.5

docs/source/intro.rst

@@ -34,6 +34,8 @@ This exposes a standard interface to talk to processes, sockets, serial ports,
 and all manner of things, along with some nifty helpers for common tasks.
 For example, remote connections via :mod:`pwnlib.tubes.remote`.
 
+::
+
     >>> conn = remote('ftp.ubuntu.com',21)
     >>> conn.recvline() # doctest: +ELLIPSIS
     b'220 ...'

extra/docker/base/Dockerfile

@@ -32,8 +32,10 @@ RUN apt-get update \
         binutils-aarch64-linux-gnu \
         binutils-mips-linux-gnu \
         binutils-mipsel-linux-gnu \
+        binutils-msp430 \
         binutils-powerpc-linux-gnu \
         binutils-powerpc64-linux-gnu \
+        binutils-s390x-linux-gnu \
         binutils-sparc64-linux-gnu \
         binutils-riscv64-linux-gnu \
         tmux \

pwn/toplevel.py

@@ -89,7 +89,15 @@ try:
 except ImportError:
     pass
 else:
-    colored_traceback.add_hook()
+    try:
+        colored_traceback.add_hook()
+    except Exception:
+        # Exception: curses.error
+        # colored_traceback (curses.setupterm()) fails if TERM is unset.
+        # This is not critical, so we just ignore it.
+        # We cannot import `curses` for `curses.error` because it is not
+        # available on all platforms (e.g. Windows).
+        pass
 
 # Equivalence with the default behavior of "from import *"
 # __all__ = [x for x in tuple(globals()) if not x.startswith('_')]

pwnlib/adb/adb.py

@@ -1275,8 +1275,28 @@ properties = Property()
 
 def _build_date():
     """Returns the build date in the form YYYY-MM-DD as a string"""
+    import datetime
+
+    # Use ro.build.date.utc (integer epoch seconds) which is set by the
+    # AOSP build system and available on all standard Android devices.
+    # This avoids ro.build.date which is locale-dependent and can contain
+    # non-ASCII characters that dateutil cannot parse.  See #2513.
+    utc = getprop('ro.build.date.utc')
+    if utc and utc.strip().isdigit():
+        try:
+            as_datetime = datetime.datetime.fromtimestamp(int(utc.strip()), dateutil.tz.UTC)
+            return as_datetime.strftime('%Y-%b-%d')
+        except (OSError, OverflowError, ValueError):
+            pass
+
+    # Fallback for non-standard builds missing ro.build.date.utc.
     as_string = getprop('ro.build.date')
-    as_datetime =  dateutil.parser.parse(as_string)
+    if not as_string:
+        return ''
+    try:
+        as_datetime = dateutil.parser.parse(as_string)
+    except (ValueError, OverflowError):
+        return as_string
     return as_datetime.strftime('%Y-%b-%d')
 
 def find_ndk_project_root(source):

pwnlib/asm.py

@@ -170,6 +170,8 @@ def which_binutils(util, check_version=False):
         '.../bin/arm-...-as'
         >>> which_binutils('as', arch='powerpc') #doctest: +ELLIPSIS
         '.../bin/powerpc...-as'
+        >>> which_binutils('as', arch='mips', endianness='little') #doctest: +ELLIPSIS
+        '.../bin/mipsel...-as'
         >>> which_binutils('as', arch='msp430') #doctest: +SKIP
         ...
         Traceback (most recent call last):
@@ -178,6 +180,12 @@ def which_binutils(util, check_version=False):
     """
     arch = context.arch
 
+    # Handle endianness-specific naming for mips/mips64
+    arch = {
+        ('mips', 'little'): 'mipsel',
+        ('mips64', 'little'): 'mips64el',
+    }.get((arch, context.endianness), arch)
+
     # Fix up pwntools vs Debian triplet naming, and account
     # for 'thumb' being its own pwntools architecture.
     arches = [arch] + {
@@ -185,11 +193,14 @@ def which_binutils(util, check_version=False):
         'i386':   ['x86_64', 'amd64'],
         'i686':   ['x86_64', 'amd64'],
         'amd64':  ['x86_64', 'i386'],
-        'mips64': ['mips'],
+        'mips': ['mipsel'],
+        'mipsel': ['mips'],
+        'mips64': ['mips', 'mipsel'],
+        'mips64el': ['mipsel', 'mips'],
         'powerpc64': ['powerpc'],
         'sparc64': ['sparc'],
-        'riscv32': ['riscv32', 'riscv64', 'riscv'],
-        'riscv64': ['riscv64', 'riscv32', 'riscv'],
+        'riscv32': ['riscv64', 'riscv'],
+        'riscv64': ['riscv32', 'riscv'],
     }.get(arch, [])
 
     # If one of the candidate architectures matches the native

pwnlib/commandline/template.py

@@ -115,6 +115,18 @@ def get_docker_image_libraries():
     return libc_basename, ld_basename
 
 def detect_missing_binaries(args):
+    """Automatically detects challenge binaries and libraries in the current directory.
+    
+    This function scans the current directory for executable files, libc, and ld libraries.
+    If a Dockerfile is present and no libraries are found, it attempts to extract them from
+    the Docker image, but only if the binary is not statically linked.
+    
+    Args:
+        args: Argument namespace containing exe and libc attributes.
+        
+    Returns:
+        tuple: A pair of (executable_path, libc_path) where either may be None if not found.
+    """
     log.info("Automatically detecting challenge binaries...")
     # look for challenge binary, libc, and ld in current directory
     exe, libc, ld = args.exe, args.libc, None
@@ -138,7 +150,19 @@ def detect_missing_binaries(args):
         elif len(other_files) > 1:
             log.warning("Failed to find challenge binary. There are multiple binaries in the current directory: %s", other_files)
     
-    if has_dockerfile and exe and not (libc or ld): 
+    # Check if the binary is statically linked before trying to extract libraries from Docker
+    is_statically_linked = False
+    if exe:
+        try:
+            binary = ELF(exe, checksec=False)
+            is_statically_linked = binary.statically_linked
+            if is_statically_linked:
+                log.info("Binary is statically linked, no need for external libraries")
+        except Exception as e:
+            log.warning("Could not check if binary is statically linked: %s", e)
+    
+    # Only extract libraries from Docker if the binary is not statically linked
+    if has_dockerfile and exe and not (libc or ld) and not is_statically_linked: 
         libc, ld = get_docker_image_libraries()
 
     if exe != args.exe:

pwnlib/context/__init__.py

@@ -359,6 +359,7 @@ class ContextType(object):
         'encoding': 'auto',
         'endian': 'little',
         'gdbinit': "",
+        'gdb_binary': "",
         'kernel': None,
         'local_libcdb': "/var/lib/libc-database",
         'log_level': logging.INFO,
@@ -1539,6 +1540,20 @@ class ContextType(object):
         """
         return str(value)
 
+    @_validator
+    def gdb_binary(self, value):
+        """Path to the binary that is used when running GDB locally.
+
+        This is useful when you have multiple versions of gdb installed or the gdb binary is
+        called something different.
+
+        If set to an empty string, pwntools will try to search for a reasonable gdb binary from 
+        the path.
+
+        Default value is ``""``.
+        """
+        return str(value)
+
     @_validator
     def cyclic_alphabet(self, alphabet):
         """Cyclic alphabet.

pwnlib/elf/corefile.py

@@ -555,6 +555,9 @@ class Corefile(ELF):
         # Pointer to the entry point
         self.at_entry = 0
 
+        # Pointer to the vdso
+        self.at_sysinfo_ehdr = None
+
         try:
             super(Corefile, self).__init__(*a, **kw)
         except IOError:
@@ -611,6 +614,8 @@ class Corefile(ELF):
 
             if not self.stack and self.mappings:
                 self.stack = self.mappings[-1].stop
+                if self.mappings[-1].start == 0xffffffffff600000 and len(self.mappings) > 1:
+                    self.stack = self.mappings[-2].stop
 
             if self.stack and self.mappings:
                 for mapping in self.mappings:
@@ -1510,7 +1515,18 @@ class CorefileFinder(object):
         # should be unique enough that we can just glob.
 
         boot_id = read('/proc/sys/kernel/random/boot_id').strip().decode()
-        path = self.exe.replace('/', '_')
+
+        # Use the absolute path of the executable
+        # Apport uses the executable's path to determine the core dump filename
+        #
+        # Reference source:
+        # https://github.com/canonical/apport/blob/4bbb179b8f92989bf7c1ee3692074f35d70ef3e8/data/apport#L110
+        # https://github.com/canonical/apport/blob/4bbb179b8f92989bf7c1ee3692074f35d70ef3e8/apport/fileutils.py#L599
+        #
+        # Apport calls `get_core_path` with `options.executable_path`, which corresponds to
+        # the executable's pathname, as specified by the `%E` placeholder
+        # in the core pattern (see `man core` and `apport --help`).
+        path = os.path.abspath(self.exe).replace('/', '_').replace('.', '_')
 
         # Format the name
         corefile_name = 'core.{path}.{uid}.{boot_id}.{pid}.*'.format(

pwnlib/elf/elf.py

@@ -715,29 +715,56 @@ class ELF(ELFFile):
 
     @property
     def libs(self):
-        """Dictionary of ``{path: address}`` for every library loaded for this ELF."""
+        """Dictionary of ``{path: address}`` for every library loaded for this ELF.
+
+        .. warning::
+
+            Getting this attribute actually runs the executable.
+            Make sure that you trust the binary you are exploiting.
+            If it adds itself as ``DT_NEEDED``, has overlapping segments,
+            ambiguous headers, or employs text relocations, it can run arbitrary
+            code even though you are just inspecting it.
+            Running is the only reliable way to ensure all the libraries are
+            loaded from the correct paths, because some of them may change
+            loading logic.
+
+            Exploitability first noticed at CWTE CTF 2025.
+        """
         if self._libs is None:
             self._populate_libraries()
         return self._libs
 
     @property
     def maps(self):
-        """Dictionary of ``{name: address}`` for every mapping in this ELF's address space."""
+        """Dictionary of ``{name: address}`` for every mapping in this ELF's address space.
+
+        .. warning::
+
+            Getting this attribute actually runs the executable.
+            Make sure that you trust the binary you are exploiting
+            (see :attr:`.ELF.libs`).
+        """
         if self._maps is None:
             self._populate_libraries()
         return self._maps
 
     @property
     def libc(self):
-        """:class:`.ELF`: If this :class:`.ELF` imports any libraries which contain ``'libc[.-]``,
+        """:class:`.ELF`: If this :class:`.ELF` imports any libraries which contain ``/libc[.-]``,
         and we can determine the appropriate path to it on the local
         system, returns a new :class:`.ELF` object pertaining to that library.
         Prints the `checksec` output of the library if it was printed for the original ELF too.
 
         If not found, the value will be :const:`None`.
+
+        .. warning::
+
+            Getting this attribute actually runs the executable.
+            Make sure that you trust the binary you are exploiting
+            (see :attr:`.ELF.libs`).
         """
         for lib in self.libs:
-            if '/libc.' in lib or '/libc-' in lib:
+            if ('/libc.' in lib or '/libc-' in lib) and lib != self.path:
                 return ELF(lib, self._print_checksec)
 
     def _populate_libraries(self):
@@ -852,6 +879,13 @@ class ELF(ELFFile):
             log.warn_once("Injected /proc/self/maps code did not execute correctly")
             return {}
 
+        # Sometimes the original binary already fail to run, for example, in case glibc mismatches.
+        try:
+            int(data.split('-', 1)[0], 16)
+        except ValueError:
+            log.warn_once("Cannot execute `%s` to get /proc/self/maps", self.path)
+            return {}
+
         # Swap in the original ELF name
         data = data.replace(path, self.path)
 
@@ -919,10 +953,9 @@ class ELF(ELFFile):
                 continue
 
             for symbol in _iter_symbols(section):
-                value = symbol.entry.st_value
-                if not value:
+                if not symbol.name or symbol.entry.st_shndx == 'SHN_UNDEF':
                     continue
-                self.symbols[symbol.name] = value
+                self.symbols[symbol.name] = symbol.entry.st_value
 
     def _populate_synthetic_symbols(self):
         """Adds symbols from the GOT and PLT to the symbols dictionary.
@@ -1791,131 +1824,136 @@ class ELF(ELFFile):
             \\* Hardware limitations are ignored.
 
         If ``READ_IMPLIES_EXEC`` is set, then `all readable pages are executable`__.
-            .. __: https://github.com/torvalds/linux/blob/v6.3/fs/binfmt_elf.c#L1008-L1009
-            .. code-block:: c
 
-                if (elf_read_implies_exec(loc->elf_ex, executable_stack))
-                    current->personality |= READ_IMPLIES_EXEC;
+            .. __: https://github.com/torvalds/linux/blob/v6.3/fs/binfmt_elf.c#L1008-L1009
+
+        .. code-block:: c
+
+            if (elf_read_implies_exec(loc->elf_ex, executable_stack))
+                current->personality |= READ_IMPLIES_EXEC;
 
         .. [#x86_5.7]
             `source <https://github.com/torvalds/linux/blob/v5.7/arch/x86/include/asm/elf.h#L285-L286>`__
 
-            .. code-block:: c
+        .. code-block:: c
 
-                #define elf_read_implies_exec(ex, executable_stack)	\\
-                    (executable_stack != EXSTACK_DISABLE_X)
+            #define elf_read_implies_exec(ex, executable_stack)	\\
+                (executable_stack != EXSTACK_DISABLE_X)
 
         .. [#x86_5.8]
             `source <https://github.com/torvalds/linux/blob/v5.8/arch/x86/include/asm/elf.h#L305-L306>`__
 
-            .. code-block:: c
+        .. code-block:: c
 
-                #define elf_read_implies_exec(ex, executable_stack)	\\
-                    (mmap_is_ia32() && executable_stack == EXSTACK_DEFAULT)
+            #define elf_read_implies_exec(ex, executable_stack)	\\
+                (mmap_is_ia32() && executable_stack == EXSTACK_DEFAULT)
 
-            `mmap_is_ia32()`__:
-                .. __: https://github.com/torvalds/linux/blob/v5.8/arch/x86/include/asm/elf.h#L318-L321
-                .. code-block:: c
+        `mmap_is_ia32()`__:
 
-                    /*
-                     * True on X86_32 or when emulating IA32 on X86_64
-                     */
-                    static inline int mmap_is_ia32(void)
+            .. __: https://github.com/torvalds/linux/blob/v5.8/arch/x86/include/asm/elf.h#L318-L321
+
+        .. code-block:: c
+
+            /*
+             * True on X86_32 or when emulating IA32 on X86_64
+             */
+            static inline int mmap_is_ia32(void)
 
         .. [#arm_5.7]
             `source <https://github.com/torvalds/linux/blob/v5.7/arch/arm/kernel/elf.c#L85-L92>`__
 
-            .. code-block:: c
+        .. code-block:: c
 
-                int arm_elf_read_implies_exec(int executable_stack)
-                {
-                    if (executable_stack != EXSTACK_DISABLE_X)
-                        return 1;
-                    if (cpu_architecture() < CPU_ARCH_ARMv6)
-                        return 1;
-                    return 0;
-                }
+            int arm_elf_read_implies_exec(int executable_stack)
+            {
+                if (executable_stack != EXSTACK_DISABLE_X)
+                    return 1;
+                if (cpu_architecture() < CPU_ARCH_ARMv6)
+                    return 1;
+                return 0;
+            }
 
         .. [#arm_5.8]
             `source <https://github.com/torvalds/linux/blob/v5.8/arch/arm/kernel/elf.c#L104-L111>`__
 
-            .. code-block:: c
+        .. code-block:: c
 
-                int arm_elf_read_implies_exec(int executable_stack)
-                {
-                    if (executable_stack == EXSTACK_DEFAULT)
-                        return 1;
-                    if (cpu_architecture() < CPU_ARCH_ARMv6)
-                        return 1;
-                    return 0;
-                }
+            int arm_elf_read_implies_exec(int executable_stack)
+            {
+                if (executable_stack == EXSTACK_DEFAULT)
+                    return 1;
+                if (cpu_architecture() < CPU_ARCH_ARMv6)
+                    return 1;
+                return 0;
+            }
 
         .. [#mips_5.17]
             `source <https://github.com/torvalds/linux/blob/v5.17/arch/mips/kernel/elf.c#L329-L342>`__
 
-            .. code-block:: c
+        .. code-block:: c
 
-                int mips_elf_read_implies_exec(void *elf_ex, int exstack)
-                {
-                    if (exstack != EXSTACK_DISABLE_X) {
-                        /* The binary doesn't request a non-executable stack */
-                        return 1;
-                    }
-                    if (!cpu_has_rixi) {
-                        /* The CPU doesn't support non-executable memory */
-                        return 1;
-                    }
-                    return 0;
+            int mips_elf_read_implies_exec(void *elf_ex, int exstack)
+            {
+                if (exstack != EXSTACK_DISABLE_X) {
+                    /* The binary doesn't request a non-executable stack */
+                    return 1;
                 }
+                if (!cpu_has_rixi) {
+                    /* The CPU doesn't support non-executable memory */
+                    return 1;
+                }
+                return 0;
+            }
 
         .. [#mips_5.18]
             `source <https://github.com/torvalds/linux/blob/v5.18/arch/mips/kernel/elf.c#L329-L336>`__
 
-            .. code-block:: c
+        .. code-block:: c
 
-                int mips_elf_read_implies_exec(void *elf_ex, int exstack)
-                {
-                    /*
-                     * Set READ_IMPLIES_EXEC only on non-NX systems that
-                     * do not request a specific state via PT_GNU_STACK.
-                     */
-                    return (!cpu_has_rixi && exstack == EXSTACK_DEFAULT);
-                }
+            int mips_elf_read_implies_exec(void *elf_ex, int exstack)
+            {
+                /*
+                 * Set READ_IMPLIES_EXEC only on non-NX systems that
+                 * do not request a specific state via PT_GNU_STACK.
+                 */
+                return (!cpu_has_rixi && exstack == EXSTACK_DEFAULT);
+            }
 
         .. [#powerpc]
             `source <https://github.com/torvalds/linux/blob/v6.3/arch/powerpc/include/asm/elf.h#L82-L108>`__
 
-            .. code-block:: c
+        .. code-block:: c
 
-                #ifdef __powerpc64__
-                /* stripped */
-                # define elf_read_implies_exec(ex, exec_stk) (is_32bit_task() ? \\
-                        (exec_stk == EXSTACK_DEFAULT) : 0)
-                #else 
-                # define elf_read_implies_exec(ex, exec_stk) (exec_stk == EXSTACK_DEFAULT)
-                #endif /* __powerpc64__ */
+            #ifdef __powerpc64__
+            /* stripped */
+            # define elf_read_implies_exec(ex, exec_stk) (is_32bit_task() ? \\
+                    (exec_stk == EXSTACK_DEFAULT) : 0)
+            #else
+            # define elf_read_implies_exec(ex, exec_stk) (exec_stk == EXSTACK_DEFAULT)
+            #endif /* __powerpc64__ */
 
         .. [#ia64]
             `source <https://github.com/torvalds/linux/blob/v6.3/arch/ia64/include/asm/elf.h#L203-L204>`__
 
-            .. code-block:: c
+        .. code-block:: c
 
-                #define elf_read_implies_exec(ex, executable_stack)					\\
-                    ((executable_stack!=EXSTACK_DISABLE_X) && ((ex).e_flags & EF_IA_64_LINUX_EXECUTABLE_STACK) != 0)
+            #define elf_read_implies_exec(ex, executable_stack)					\\
+                ((executable_stack!=EXSTACK_DISABLE_X) && ((ex).e_flags & EF_IA_64_LINUX_EXECUTABLE_STACK) != 0)
 
-            EF_IA_64_LINUX_EXECUTABLE_STACK__:
-                .. __: https://github.com/torvalds/linux/blob/v6.3/arch/ia64/include/asm/elf.h#L33
+        EF_IA_64_LINUX_EXECUTABLE_STACK__:
 
-                .. code-block:: c
+            .. __: https://github.com/torvalds/linux/blob/v6.3/arch/ia64/include/asm/elf.h#L33
 
-                    #define EF_IA_64_LINUX_EXECUTABLE_STACK	0x1	/* is stack (& heap) executable by default? */
+        .. code-block:: c
+
+            #define EF_IA_64_LINUX_EXECUTABLE_STACK	0x1	/* is stack (& heap) executable by default? */
 
         .. [#the_rest]
             `source <https://github.com/torvalds/linux/blob/v6.3/include/linux/elf.h#L13>`__
 
-            .. code-block:: c
+        .. code-block:: c
 
-                # define elf_read_implies_exec(ex, have_pt_gnu_stack)	0
+            # define elf_read_implies_exec(ex, have_pt_gnu_stack)	0
         """
         if not self.executable:
             return True

pwnlib/elf/plt.py

@@ -65,7 +65,7 @@ def __ensure_memory_to_run_unicorn():
         from mmap import mmap, MAP_ANON, MAP_PRIVATE, PROT_EXEC, PROT_READ, PROT_WRITE
 
         mm = mmap(
-            -1, 1024 * 1024 * 1024, MAP_PRIVATE | MAP_ANON, PROT_WRITE | PROT_READ | PROT_EXEC
+            -1, 1024 * 1024 * 1024, MAP_PRIVATE | MAP_ANON, PROT_WRITE | PROT_READ
         )
         mm.close()
     except OSError:
@@ -78,6 +78,13 @@ def __ensure_memory_to_run_unicorn():
 def prepare_unicorn_and_context(elf, got, address, data):
     import unicorn as U
 
+    if elf.arch in ('mips', 'mips64'):
+        # Unicorn 2.1.3 and 2.1.4 have a crashing bug when trying to emulate MIPS.
+        # https://github.com/unicorn-engine/unicorn/issues/2134
+        if U.UC_VERSION_MAJOR == 2 and U.UC_VERSION_MINOR == 1 and U.UC_VERSION_PATCH in (3, 4):
+            raise OSError("Unicorn Engine 2.1.3 and 2.1.4 have a crashing bug when emulating MIPS, please install an earlier or later version: https://github.com/unicorn-engine/unicorn/issues/2134")
+
+
     __ensure_memory_to_run_unicorn()
 
     # Instantiate the emulator with the correct arguments for the current

pwnlib/filepointer.py

@@ -156,6 +156,18 @@ class FileStructure(object):
          _wide_data: 0xdeadbeef
          unknown2: 0x0
          vtable: 0x0}
+
+        Bytes fields are padded to the correct field size.
+        For example, unknown2 is 56 bytes on i386, so a short value gets
+        zero-padded to 56, not to context.bytes (4):
+
+        >>> context.clear(arch='i386')
+        >>> fileStr2 = FileStructure(null=0)
+        >>> fileStr2.vtable = 0x561859f0
+        >>> old_len = len(bytes(fileStr2))
+        >>> fileStr2.unknown2 = b'AB'
+        >>> len(bytes(fileStr2)) == old_len
+        True
     """
 
     vars_=[]
@@ -186,7 +198,7 @@ class FileStructure(object):
         structure = b''
         for val in self.vars_:
             if isinstance(getattr(self, val), bytes):
-                structure += getattr(self, val).ljust(context.bytes, b'\x00')
+                structure += getattr(self, val).ljust(self.length[val], b'\x00')
             else:
                 if self.length[val] > 0:
                     structure += pack(getattr(self, val), self.length[val]*8)
@@ -215,7 +227,7 @@ class FileStructure(object):
         structure = b''
         for val in self.vars_:
             if isinstance(getattr(self, val), bytes):
-                structure += getattr(self, val).ljust(context.bytes, b'\x00')
+                structure += getattr(self, val).ljust(self.length[val], b'\x00')
             else:
                 structure += pack(getattr(self, val), self.length[val]*8)
             if val == v:

pwnlib/gdb.py

@@ -143,7 +143,6 @@ from __future__ import division
 
 import os
 import platform
-import psutil
 import random
 import re
 import six
@@ -353,10 +352,6 @@ def _gdbserver_args(pid=None, path=None, port=0, gdbserver_args=None, args=None,
         gdbserver_args += ['--wrapper', python_wrapper_script, '--']
     elif env is not None:
         gdbserver_args += ['--wrapper', which('env'), '-i'] + env_args + ['--']
-    # --no-startup-with-shell is required for forking shells like SHELL=/bin/fish
-    # https://github.com/Gallopsled/pwntools/issues/2377
-    else:
-        gdbserver_args += ['--no-startup-with-shell']
 
     gdbserver_args += ['localhost:%d' % port]
     gdbserver_args += args
@@ -750,6 +745,12 @@ def binary():
         >>> gdb.binary() # doctest: +SKIP
         '/usr/bin/gdb'
     """
+    if context.gdb_binary:
+        gdb = misc.which(context.gdb_binary)
+        if not gdb:
+            log.warn_once('Path to gdb binary `{}` not found'.format(context.gdb_binary))
+        return gdb
+
     gdb = misc.which('pwntools-gdb') or misc.which('gdb')
 
     if not context.native:
@@ -1520,28 +1521,33 @@ def corefile(process):
                       'contain all of the data in the process.\n'
                       'Upgrade to GDB >= 7.11 for better core-dumps.' % binary())
 
-    # This is effectively the same as what the 'gcore' binary does
-    gdb_args = ['-batch',
-                '-q',
-                '-nx',
-                '-ex', 'set pagination off',
-                '-ex', 'set height 0',
-                '-ex', 'set width 0',
-                '-ex', 'set use-coredump-filter on',
-                '-ex', 'generate-core-file %s' % corefile_path,
-                '-ex', 'detach']
+    # This is effectively the same as what the 'gcore' binary does.
+    # Run GDB directly instead of going through attach(), which has to
+    # race GDB to observe the tracer before it already detached again.
+    cmd = [binary(),
+           '-batch',
+           '-q',
+           '-nx',
+           '-ex', 'set pagination off',
+           '-ex', 'set height 0',
+           '-ex', 'set width 0',
+           '-ex', 'set use-coredump-filter on',
+           '-ex', 'attach %d' % process.pid,
+           '-ex', 'generate-core-file %s' % corefile_path,
+           '-ex', 'detach']
 
-    with context.local(terminal = ['sh', '-c']):
-        with context.quiet:
-            pid = attach(process, gdb_args=gdb_args)
-            log.debug("Got GDB pid %d", pid)
-            try:
-                psutil.Process(pid).wait()
-            except psutil.Error:
-                pass
+    with context.quiet:
+        gdb = tubes.process.process(cmd, stdout=tubes.process.PIPE)
+        gdb.wait_for_close(timeout=60)
+        timed_out = gdb.poll() is None
+        output = gdb.recvallS(timeout=0)
+
+    if timed_out:
+        log.error("GDB did not finish dumping process %d within 60 seconds", process.pid)
 
     if not os.path.exists(corefile_path):
-        log.error("Could not generate a corefile for process %d", process.pid)
+        log.error("Could not generate a corefile for process %d (maybe check /proc/sys/kernel/yama/ptrace_scope)\n%s",
+                  process.pid, output)
 
     return elf.corefile.Core(corefile_path)
 

pwnlib/lexer.py

@@ -95,7 +95,7 @@ class PwntoolsLexer(RegexLexer):
         'whitespace': [
             (r'\n', Text),
             (r'\s+', Text),
-            (r'/\*.*?\*/', Comment),
+            (r'(?ms)/\*.*?\*/', Comment),
             (r';.*$', Comment)
         ],
         'punctuation': [

pwnlib/libcdb.py

@@ -9,6 +9,7 @@ import time
 import six
 import tempfile
 import struct
+import sys
 
 from pwnlib.context import context
 from pwnlib.elf import ELF
@@ -65,6 +66,8 @@ MAP_TYPES = {
 }
 
 DEBUGINFOD_SERVERS = [
+    'https://debuginfod.ubuntu.com/',
+    'https://debuginfod.debian.net/',
     'https://debuginfod.elfutils.org/',
 ]
 
@@ -293,49 +296,6 @@ def search_by_hash(search_target, search_type='build_id', unstrip=True, offline_
 
     return cache
 
-def _search_debuginfo_by_hash(base_url, hex_encoded_id):
-    """
-    Given a hex-encoded build_id, attempt to download a matching debuginfo from the debuginfod server.
-
-    >>> debuginfo_file = _search_debuginfo_by_hash(DEBUGINFOD_SERVERS[0], 'd1704d25fbbb72fa95d517b883131828c0883fe9')
-    >>> debuginfo_file is not None
-    True
-    >>> 'main_arena' in ELF(debuginfo_file).symbols
-    True
-    """
-    # Deferred import because it's slow
-    import requests
-    from six.moves import urllib
-
-    # Check if we tried this buildid before.
-    cache, cache_valid = _check_elf_cache('libcdb_dbg', hex_encoded_id, 'build_id')
-    if cache_valid:
-        return cache
-    
-    # We searched for this buildid before, but didn't find anything.
-    if cache is None:
-        return None
-
-    # Try to find separate debuginfo.
-    url  = '/buildid/{}/debuginfo'.format(hex_encoded_id)
-    url  = urllib.parse.urljoin(base_url, url)
-    data = b""
-    log.debug("Downloading data from debuginfod: %s", url)
-    try:
-        data = wget(url, timeout=20)
-    except requests.RequestException as e:
-        log.warn_once("Failed to fetch libc debuginfo for build_id %s from %s: %s", hex_encoded_id, base_url, e)
-    
-    # Save whatever we got to the cache
-    write(cache, data or b'')
-
-    # Return ``None`` if we did not get a valid ELF file
-    if not data or not data.startswith(b'\x7FELF'):
-        log.warn_once("Could not fetch libc debuginfo for build_id %s from %s", hex_encoded_id, base_url)
-        return None
-
-    return cache
-
 def _check_elf_cache(cache_type, search_target, search_type):
     """
     Check if there already is an ELF file for this hash in the cache.
@@ -394,15 +354,15 @@ def unstrip_libc(filename):
         >>> libc = ELF(filename)
         >>> 'main_arena' in libc.symbols
         False
-        >>> unstrip_libc(filename)
+        >>> unstrip_libc(filename) # doctest: +SKIP
         True
         >>> libc = ELF(filename)
-        >>> hex(libc.symbols.main_arena)
+        >>> hex(libc.symbols.main_arena) # doctest: +SKIP
         '0x219c80'
         >>> unstrip_libc(pwnlib.data.elf.get('test-x86'))
         False
         >>> filename = search_by_build_id('d1704d25fbbb72fa95d517b883131828c0883fe9', unstrip=True)
-        >>> 'main_arena' in ELF(filename).symbols
+        >>> 'main_arena' in ELF(filename).symbols # doctest: +SKIP
         True
     """
     if not which('eu-unstrip'):
@@ -420,16 +380,45 @@ def unstrip_libc(filename):
 
     log.debug('Trying debuginfod servers: %r', DEBUGINFOD_SERVERS)
 
-    for server_url in DEBUGINFOD_SERVERS:
-        libc_dbg = _search_debuginfo_by_hash(server_url, enhex(libc.buildid))
-        if libc_dbg:
-            break
-    else:
-        log.warn_once('Couldn\'t find debug info for libc with build_id %s on any debuginfod server.', enhex(libc.buildid))
-        return False
+    # Deferred import because it's slow
+    import requests
+    from six.moves import urllib
+
+    hex_encoded_id = enhex(libc.buildid)
+
+    # Check if we tried this buildid before.
+    cache, cache_valid = _check_elf_cache('libcdb_dbg', hex_encoded_id, 'build_id')
+    if not cache_valid:
+        # Cached negative result so we don't look for missing debug info everytime.
+        if cache is None:
+            return False
+        else:
+            for server_url in DEBUGINFOD_SERVERS:
+                # Try to find separate debuginfo.
+                url  = '/buildid/{}/debuginfo'.format(hex_encoded_id)
+                url  = urllib.parse.urljoin(server_url, url)
+                data = b""
+                log.debug("Downloading data from debuginfod: %s", url)
+                try:
+                    data = wget(url, timeout=20)
+
+                    # Try next server if we didn't get a valid ELF file
+                    if not data or not data.startswith(b'\x7FELF'):
+                        log.warn_once("Could not fetch libc debuginfo for build_id %s from %s", hex_encoded_id, server_url)
+                        continue
+                    break
+                except requests.RequestException as e:
+                    log.warn_once("Failed to fetch libc debuginfo for build_id %s from %s: %s", hex_encoded_id, server_url, e)
+            else:
+                write(cache, data or b'')
+                log.warn_once('Couldn\'t find debug info for libc with build_id %s on any debuginfod server.', enhex(libc.buildid))
+                return False
+            
+            # Save whatever we got to the cache
+            write(cache, data or b'')
 
     # Add debug info to given libc binary inplace.
-    p = process(['eu-unstrip', '-o', filename, filename, libc_dbg])
+    p = process(['eu-unstrip', '-o', filename, filename, cache])
     output = p.recvall()
     p.close()
 
@@ -498,7 +487,7 @@ def _extract_tarfile(cache_dir, data_filename, tarball):
 
 def _extract_debfile(cache_dir, package_filename, package):
     # Extract data.tar in the .deb archive.
-    if six.PY2:
+    if sys.version_info < (3, 6):
         if not which('ar'):
             log.error('Missing command line tool "ar" to extract .deb archive. Please install "ar" first.')
 

pwnlib/regsort.py

@@ -15,7 +15,7 @@ from pwnlib.log import getLogger
 
 log = getLogger(__name__)
 
-def check_cycle(reg, assignments):
+def check_cycle(reg, assignments, mapping=None):
     """Walk down the assignment list of a register,
     return the path walked if it is encountered again.
 
@@ -37,31 +37,38 @@ def check_cycle(reg, assignments):
         >>> check_cycle('a', {'a': 'b', 'b': 'c', 'c': 'd', 'd': 'a'})
         ['a', 'b', 'c', 'd']
     """
-    return check_cycle_(reg, assignments, [])
+    if mapping is None:
+        mapping = {}
+    return check_cycle_(reg, assignments, [], mapping)
 
-def check_cycle_(reg, assignments, path):
-    target = assignments[reg]
+def check_cycle_(reg, assignments, path, mapping):
+    target = assignments.get(reg)
+    if target is None:
+        real_reg = mapping.get(reg, reg)
+        reg, target = next((k, v) for k,v in assignments.items() if mapping.get(k, k) == real_reg)
     path.append(reg)
 
+    real_target = mapping.get(target, target)
+
     # No cycle, some other value (e.g. 1)
-    if target not in assignments:
+    if all(real_target != mapping.get(k, k) for k in assignments):
         return []
 
     # Found a cycle
-    if target in path:
+    if any(real_target == mapping.get(k, k) for k in path):
         # Does the cycle *start* with target?
         # This determines whether the original register is
         # in the cycle, or just depends on registers in one.
-        if target == path[0]:
+        if real_target == mapping.get(path[0], path[0]):
             return path
 
         # Just depends on one.
         return []
 
     # Recurse
-    return check_cycle_(target, assignments, path)
+    return check_cycle_(target, assignments, path, mapping)
 
-def extract_dependencies(reg, assignments):
+def extract_dependencies(reg, assignments, mapping=None):
     """Return a list of all registers which directly
     depend on the specified register.
 
@@ -77,7 +84,9 @@ def extract_dependencies(reg, assignments):
         ['b', 'c']
     """
     # sorted() is only for determinism
-    return sorted([k for k,v in assignments.items() if v == reg])
+    if mapping is None:
+        mapping = {}
+    return sorted([k for k,v in assignments.items() if mapping.get(v, v) == mapping.get(reg, reg)])
 
 
 def resolve_order(reg, deps):
@@ -110,7 +119,7 @@ def depends_on_cycle(reg, assignments, in_cycles):
         reg = assignments.get(reg, None)
     return False
 
-def regsort(in_out, all_regs, tmp = None, xchg = True, randomize = None):
+def regsort(in_out, all_regs, mapping = None, tmp = None, xchg = True, randomize = None):
     """
     Sorts register dependencies.
 
@@ -181,6 +190,15 @@ def regsort(in_out, all_regs, tmp = None, xchg = True, randomize = None):
 
         >>> regsort({'a': 1, 'b': 2}, R)
         [('mov', 'a', 1), ('mov', 'b', 2)]
+        >>> M = {'ax': 'eax', 'eax': 'eax', 'ebx': 'ebx'}
+
+        Registers which are assigned one of their own aliases
+        are dropped, like same-name self-assignments.
+
+        >>> regsort({'ax': 'eax'}, M)
+        []
+        >>> regsort({'ax': 'eax'}, M, xchg=False)
+        []
         >>> regsort({'a': 'b', 'b': 'a'}, R)
         [('xchg', 'a', 'b')]
         >>> regsort({'a': 'b', 'b': 'a'}, R, tmp='X') #doctest: +NORMALIZE_WHITESPACE
@@ -233,16 +251,24 @@ def regsort(in_out, all_regs, tmp = None, xchg = True, randomize = None):
     if randomize is None:
         randomize = context.randomize
 
+    if mapping is None:
+        if hasattr(all_regs, 'keys'):
+            mapping = all_regs
+        else:
+            mapping = {}
+
     sentinel = object()
 
     # Drop all registers which will be set to themselves.
     #
     # For example, {'eax': 'eax'}
-    in_out = {k:v for k,v in in_out.items() if k != v}
+    # This must go through the mapping, so that {'ax': 'eax'} is
+    # recognized as a self-assignment as well.
+    in_out = {k:v for k,v in in_out.items() if mapping.get(k, k) != mapping.get(v, v)}
 
     # Collapse constant values
     #
-    # For eaxmple, {'eax': 0, 'ebx': 0} => {'eax': 0, 'ebx': 'eax'}
+    # For example, {'eax': 0, 'ebx': 0} => {'eax': 0, 'ebx': 'eax'}
     v_k = defaultdict(list)
     for k,v in sorted(in_out.items()):
         if v not in all_regs and v != 0:
@@ -263,7 +289,9 @@ def regsort(in_out, all_regs, tmp = None, xchg = True, randomize = None):
     # which are also 'outputs'.
     #
     # For example, {'eax': 1, 'ebx': 2, 'ecx': 'edx'}
-    if not any(v in in_out for k,v in in_out.items()):
+    inps = {mapping.get(k, k) for k in in_out}
+    outs = {mapping.get(v, v) for v in in_out.values()}
+    if not inps & outs:
         result = [('mov', k,in_out[k]) for k in sorted(in_out)]
 
         if randomize:
@@ -280,7 +308,7 @@ def regsort(in_out, all_regs, tmp = None, xchg = True, randomize = None):
     # Output:  {'A': [], 'B': ['A', 'C'], 'C': []}
     #
     # In this case, both A and C must be set before B.
-    deps  = {r: extract_dependencies(r, in_out) for r in in_out}
+    deps  = {r: extract_dependencies(r, in_out, mapping) for r in in_out}
 
     # Final result which will be returned
     result = []
@@ -299,7 +327,7 @@ def regsort(in_out, all_regs, tmp = None, xchg = True, randomize = None):
 
     while cycle_candidates:
         reg   = cycle_candidates[0]
-        cycle = check_cycle(reg, in_out)
+        cycle = check_cycle(reg, in_out, mapping)
 
         if cycle:
             if randomize:
@@ -395,6 +423,10 @@ def regsort(in_out, all_regs, tmp = None, xchg = True, randomize = None):
 
     if tmp:
         for cycle in cycles:
+            if len(cycle) == 1:
+                [reg] = cycle
+                result.append(('mov', reg, in_out[reg]))
+                continue
 
             first = cycle[0]
             last  = cycle[-1]
@@ -411,6 +443,9 @@ def regsort(in_out, all_regs, tmp = None, xchg = True, randomize = None):
     else:
         for cycle in cycles:
             size = len(cycle)
+            if size == 1:
+                [reg] = cycle
+                result.append(('mov', reg, in_out[reg]))
             for i in range(size-1):
                 result.append(('xchg', cycle[i], cycle[(i+1) % size]))
 

pwnlib/rop/rop.py

@@ -841,7 +841,7 @@ class ROP(object):
         """
         if isinstance(object, enums):
             return str(object)
-        if isinstance(object, six.integer_types):
+        if isinstance(object, six.integer_types) and object:
             return self.unresolve(object)
         if isinstance(object, (bytes, six.text_type)):
             return repr(object)

pwnlib/shellcraft/registers.py

@@ -116,10 +116,10 @@ i386_ordered = [
     ['rbx', 'ebx', 'bx', 'bl'],
     ['rcx', 'ecx', 'cx', 'cl'],
     ['rdx', 'edx', 'dx', 'dl'],
-    ['rdi', 'edi', 'di'],
-    ['rsi', 'esi', 'si'],
-    ['rbp', 'ebp', 'bp'],
-    ['rsp', 'esp', 'sp'],
+    ['rdi', 'edi', 'di', 'dil'],
+    ['rsi', 'esi', 'si', 'sil'],
+    ['rbp', 'ebp', 'bp', 'bpl'],
+    ['rsp', 'esp', 'sp', 'spl'],
     ['r8', 'r8d', 'r8w', 'r8b'],
     ['r9', 'r9d', 'r9w', 'r9b'],
     ['r10', 'r10d', 'r10w', 'r10b'],

pwnlib/shellcraft/templates/aarch64/memcpy.asm

@@ -16,4 +16,4 @@ ${memcpy_loop}:
     ldrb w3, [x1], #1
     strb w3, [x0], #1
     subs x2, x2, #1
-    bge ${memcpy_loop}
+    bgt ${memcpy_loop}

pwnlib/shellcraft/templates/amd64/mov.asm

@@ -51,8 +51,8 @@ Example:
         xor eax, eax
         mov ax, 0xc0c0
     >>> print(shellcraft.amd64.mov('rdi', 0xff).rstrip())
-        mov edi, 0x1010101 /* 255 == 0xff */
-        xor edi, 0x10101fe
+        xor edi, edi
+        mov dil, 0xff
     >>> print(shellcraft.amd64.mov('rax', 0xdead00ff).rstrip())
         mov eax, 0x1010101 /* 3735879935 == 0xdead00ff */
         xor eax, 0xdfac01fe

pwnlib/shellcraft/templates/amd64/setregs.asm

@@ -55,7 +55,7 @@ if isinstance(eax, six.integer_types) and isinstance(edx, six.integer_types) and
     cdq = True
     reg_context.pop('rdx')
 
-sorted_regs = regsort(reg_context, registers.amd64)
+sorted_regs = regsort(reg_context, registers.amd64, registers.native64)
 %>
 % if not sorted_regs:
   /* setregs noop */

pwnlib/shellcraft/templates/amd64/windows/getexport.asm

@@ -1,24 +1,30 @@
 <%
-    from pwnlib.shellcraft import amd64, pretty
+    from pwnlib.shellcraft import amd64, common, pretty
     from pwnlib.util.packing import u64, _need_bytes
 %>
 <%docstring>Find the address of an exported function in a dll
 by manually iterating through the PE export table.
 
+The dll must be `b'kernel32.dll'` or `b'ntdll.dll'` at the moment.
+Behavior is undefined if the function is not found.
+
 Args:
-    function_name (str): The name of the function to find.
-    dll(str): The name of the DLL to find the function in.
+    function_name (bytes): The name of the function to find.
+    dll(bytes): The name of the DLL to find the function in.
     dest (str): The register to load the function address into.
 </%docstring>
-<%page args="function_name,dll='kernel32.dll',dest='rax'"/>
+<%page args="function_name,dll=b'kernel32.dll',dest='rax'"/>
 <%
 function_name = _need_bytes(function_name)
 dll = _need_bytes(dll)
-if len(function_name) > 8:
-    raise ValueError('function_name must be <= 8 bytes')
-assert dll == b'kernel32.dll'
+assert dll in (b'kernel32.dll', b'ntdll.dll')
+find_function_loop = common.label('find_function').strip()
 %>
+% if dll == b'kernel32.dll':
     ${amd64.windows.kernel32base('rbx')} /* rbx = kernel32.dll PE base */
+% elif dll == b'ntdll.dll':
+    ${amd64.windows.ntdllbase('rbx')} /* rbx = ntdll.dll PE base */
+% endif
     mov r8d, [rbx + 0x3c]
     mov rdx, r8
     add rdx, rbx
@@ -26,25 +32,42 @@ assert dll == b'kernel32.dll'
     add rdx, r9
     mov r8d, [rdx]
     add r8, rbx /* r8 = export table */
-    mov esi, [r8 + 0x20]
-    add rsi, rbx /* rsi = names table */
-    xor rcx, rcx
+    mov edx, [r8 + 0x20]
+    add rdx, rbx /* rdx = names table */
 % if len(function_name) <= 8:
     mov r9, ${pretty(u64(function_name.ljust(8, b'\x00')))}
 % else:
     ${amd64.pushstr(function_name)}
     mov r9, rsp
 % endif
+    push r8
+    xor r8, r8
 
     /* Loop through the names table */
-    FindFunction:
-        inc rcx
-        mov eax, [rsi + rcx * 4]
+    ${find_function_loop}:
+        inc r8
+        mov eax, [rdx + r8 * 4]
         add rax, rbx
-        ## TODO: implement strcmp properly for function names > 8 bytes
+        ## strcmp
+        % if len(function_name) <= 8:
         cmp qword ptr [rax], r9
-        jnz FindFunction
+        % else:
+        ${amd64.setregs({
+            'rdi': 'rax',
+            'rsi': 'r9',
+            'rcx': len(function_name),
+        })}
+        repe cmpsb
+        movzx rax, byte ptr [rsi-1]
+        movzx rcx, byte ptr [rdi-1]
+        sub rax, rcx
+        % endif
+        jnz ${find_function_loop}
 
+    ## Assume we find the function, we'll crash walking past the
+    ## end of the names table otherwise.
+    mov rcx, r8
+    pop r8
     mov esi, [r8 + 0x24]
     add rsi, rbx /* rsi = ordinals table */
     mov cx, [rsi + rcx * 2]
@@ -52,4 +75,4 @@ assert dll == b'kernel32.dll'
     add rsi, rbx /* rsi = address table */
     mov eax, [rsi + rcx * 4]
     add rax, rbx /* rax = function address */
-    mov ${dest}, rax
+    ${amd64.mov(dest, 'rax')}

pwnlib/shellcraft/templates/amd64/windows/getprocaddress.asm

@@ -11,7 +11,7 @@ Args:
     dll(str): The name of the DLL to find the function in.
     dest (str): The register to load the function address into.
 </%docstring>
-<%page args="function_name,dll='kernel32.dll',dest='rax'"/>
+<%page args="function_name,dll=b'kernel32.dll',dest='rax'"/>
 <%
 function_name = _need_bytes(function_name)
 dll = _need_bytes(dll)
@@ -25,4 +25,4 @@ assert dll == b'kernel32.dll'
     sub rsp, 0x30
     call rdi
     add rsp, ${pretty(0x30+align(8, len(function_name)))}
-    mov ${dest}, rax
+    ${amd64.mov(dest, 'rax')}

pwnlib/shellcraft/templates/arm/memcpy.asm

@@ -16,4 +16,4 @@ ${memcpy_loop}:
     ldrb r3, [r1], #1
     strb r3, [r0], #1
     subs r2, r2, #1
-    bge ${memcpy_loop}
+    bgt ${memcpy_loop}

pwnlib/shellcraft/templates/i386/mov.asm

@@ -58,12 +58,9 @@ Example:
     >>> print(shellcraft.i386.mov('eax', 0xdead00ff).rstrip())
         mov eax, -0xdead00ff
         neg eax
-    >>> print(shellcraft.i386.mov('eax', 0xc0).rstrip())
-        xor eax, eax
-        mov al, 0xc0
     >>> print(shellcraft.i386.mov('edi', 0xc0).rstrip())
-        mov edi, -0xc0
-        neg edi
+        xor edi, edi
+        mov dil, 0xc0
     >>> print(shellcraft.i386.mov('eax', 0xc000).rstrip())
         xor eax, eax
         mov ah, 0xc000 >> 8

pwnlib/shellcraft/templates/i386/setregs.asm

@@ -54,7 +54,7 @@ sorted_regs = regsort(reg_context, registers.i386)
 % if not sorted_regs:
   /* setregs noop */
 % else:
-% for how, src, dst in regsort(reg_context, registers.i386):
+% for how, src, dst in regsort(reg_context, registers.i386, registers.native32):
 % if how == 'xchg':
     xchg ${src}, ${dst}
 % else:

pwnlib/shellcraft/templates/thumb/memcpy.asm

@@ -16,4 +16,4 @@ ${memcpy_loop}:
     ldrb r3, [r1], #1
     strb r3, [r4], #1
     subs r2, r2, #1
-    bge ${memcpy_loop}
+    bgt ${memcpy_loop}

pwnlib/term/readline.py

@@ -380,7 +380,7 @@ def readline(_size=-1, prompt='', float=True, priority=10):
     from pwnlib.term import term_mode
     if not term_mode:
         six.print_(prompt, end='', flush=True)
-        return getattr(sys.stdin, 'buffer', sys.stdin).readline(_size).rstrip(b'\n')
+        return force_to_bytes(getattr(sys.stdin, 'buffer', sys.stdin).readline(_size)).rstrip(b'\n')
     show_suggestions = False
     eof = False
     if prompt:

pwnlib/tubes/process.py

@@ -1259,6 +1259,9 @@ class process(tube):
 
             >>> from pwn import *
             >>> p = process(['cat'])
+            >>> p.send(b'meow')
+            >>> p.recvuntil(b'meow')
+            b'meow'
             >>> libc_size = p.lib_size(p.libc.path)
             >>> hex(libc_size) # doctest: +SKIP
             '0x1d5000'
@@ -1325,8 +1328,7 @@ class process(tube):
         by the process to the address it is loaded at in the process' address
         space.
         """
-        from pwnlib.util.proc import memory_maps
-        maps_raw = memory_maps(self.pid)
+        maps_raw = self.poll() is None and self.maps()
 
         if not maps_raw:
             import pwnlib.elf.elf
@@ -1335,23 +1337,15 @@ class process(tube):
                 return pwnlib.elf.elf.ELF(self.executable).maps
 
         # Enumerate all of the libraries actually loaded right now.
-        maps = {}
+        libs = {}
         for mapping in maps_raw:
             path = mapping.path
             if os.sep not in path: continue
             path = os.path.realpath(path)
-            if path not in maps:
-                maps[path]=0
+            if path not in libs:
+                libs[path] = mapping.addr
 
-        for lib in maps:
-            path = os.path.realpath(lib)
-            for mapping in maps_raw:
-                if mapping.path == path:
-                    address = mapping.addr.split('-')[0]
-                    maps[lib] = int(address, 16)
-                    break
-
-        return maps
+        return libs
 
     @property
     def libc(self):
@@ -1364,7 +1358,12 @@ class process(tube):
         Example:
 
         >>> p = process("/bin/cat")
+        >>> p.send(b"meow")
+        >>> p.recvuntil(b"meow")
+        b'meow'
         >>> libc = p.libc
+        >>> libc is not None
+        True
         >>> libc # doctest: +SKIP
         ELF('/lib64/libc-...so')
         >>> p.close()

pwnlib/tubes/server.py

@@ -127,7 +127,7 @@ class server(sock):
                         return
 
                 self.rhost, self.rport = rhost[:2]
-                r = remote(self.rhost, self.rport, sock = sock)
+                r = remote(self.rhost, self.rport, sock = sock, level = self.level)
                 h.success('Got connection from %s on port %d' % (self.rhost, self.rport))
                 if callback:
                     if not blocking:

pwnlib/tubes/ssh.py

@@ -1359,7 +1359,11 @@ from ctypes import *; libc = CDLL('libc.so.6'); print(libc.getenv(%r))
             update(len(data), total)
 
         result = c.wait()
-        if result != 0:
+
+        if result == -1:
+            self.warn_once("Could not verify success of file download %r, no error code" % (remote))
+
+        if result != 0 and result != -1:
             h.failure('Could not download file %r (%r)' % (remote, result))
             return
 
@@ -1539,7 +1543,11 @@ from ctypes import *; libc = CDLL('libc.so.6'); print(libc.getenv(%r))
             s.shutdown('send')
             data   = s.recvall()
             result = s.wait()
-            if result != 0:
+
+            if result == -1:
+                self.warn_once("Could not verify success of file upload %r, no error code" % (remote))
+
+            if result != 0 and result != -1:
                 self.error("Could not upload file %r (%r)\n%s" % (remote, result, data))
 
     def upload_file(self, filename, remote = None):

pwnlib/util/packing.py

@@ -988,6 +988,11 @@ def dd(dst, src, count = 0, skip = 0, seek = 0, truncate = False):
         ('H', 'e', 'l', 'l', 'o', b'?')
         >>> dd(list('Hello!'), (63,), skip = 5)
         ['H', 'e', 'l', 'l', 'o', b'?']
+        >>> buf = bytearray(b'Hello!')
+        >>> dd(buf, b'?', skip = 5)
+        bytearray(b'Hello?')
+        >>> buf
+        bytearray(b'Hello?')
         >>> _ = open('/tmp/foo', 'w').write('A' * 10)
         >>> dd(open('/tmp/foo'), open('/dev/zero'), skip = 3, count = 4).read()
         'AAA\\x00\\x00\\x00\\x00AAA'
@@ -1105,7 +1110,12 @@ def dd(dst, src, count = 0, skip = 0, seek = 0, truncate = False):
         dst = real_dst
 
     elif isinstance(dst, (list, bytearray)):
-        dst[skip : skip + len(src)] = list(map(p8, bytearray(src)))
+        # A bytearray slice takes a byte string, while a list slice takes the
+        # single-byte bytes objects that p8() returns.
+        if isinstance(dst, bytearray):
+            dst[skip : skip + len(src)] = src
+        else:
+            dst[skip : skip + len(src)] = list(map(p8, bytearray(src)))
         if truncate:
             while len(dst) > truncate:
                 dst.pop()

pwnlib/version.py

@@ -1 +1 @@
-__version__ = '4.15.0beta0'
+__version__ = '4.15.0'

pyproject.toml

@@ -56,7 +56,7 @@ dependencies = [
     "colored_traceback<0.4; python_version < '3'",
     "colored_traceback; python_version >= '3'",
     "pathlib2; python_version < '3.4'",
-    "unix-ar; python_version >= '3'",
+    "unix-ar; python_version >= '3.6'",
     "zstandard",
 ]
 

setup.py

@@ -87,7 +87,7 @@ if not os.path.exists(PythonH):
     sys.exit(-1)
 
 setup(
-    version              = '4.15.0beta0',
+    version              = '4.15.0',
     data_files           = [('pwntools-doc',
                              glob.glob('*.md') + glob.glob('*.txt')),
                             ],

travis/libcdb_nginx_cache.conf

@@ -15,6 +15,7 @@ http {
 
         location / {
             proxy_set_header Host debuginfod.elfutils.org;
+            proxy_ssl_server_name on;
             proxy_cache_revalidate on;
             proxy_cache_key $scheme://$host$uri$is_args$query_string;
             proxy_cache_valid 200 404 12w;
@@ -29,6 +30,7 @@ http {
 
         location / {
             proxy_set_header Host libc.rip;
+            proxy_ssl_server_name on;
             proxy_cache_methods GET HEAD POST;
             proxy_cache_revalidate on;
             proxy_cache_key $scheme://$host$uri$is_args$query_string$request_body;
@@ -66,4 +68,19 @@ http {
             proxy_pass https://gitlab.com/;
         }
     }
+
+    server {
+        listen 3004;
+        proxy_cache my_cache;
+
+        location / {
+            proxy_set_header Host debuginfod.ubuntu.com;
+            proxy_ssl_server_name on;
+            proxy_cache_revalidate on;
+            proxy_cache_key $scheme://$host$uri$is_args$query_string;
+            proxy_cache_valid 200 404 12w;
+            proxy_cache_use_stale error timeout http_500 http_502 http_503 http_504 http_429;
+            proxy_pass https://debuginfod.ubuntu.com/;
+        }
+    }
 }