Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 40 additions & 0 deletions Framework/Core/COOKBOOK.md
Original file line number Diff line number Diff line change
Expand Up @@ -549,3 +549,43 @@ Streams can be explicitly enabled or disabled in code using the `O2_SIGNPOST_ENA
If a process is already running and you wish to enable one or more of its signposts logs, you can do so using the `o2-log` utility, passing the address of the log to enable and the PID of the running process. E.g. `o2-log -p <PID> -a <hook address of the signpost>`.

Finally, on macOS, you can also use Instruments to visualise your Signpost, just like any other macOS application. In order to do so you need to enable the "Signpost" instrument, making sure you add `ch.cern.aliceo2.completion` to the list of loggers to watch.

## Improving lldb experience

You can make lldb understand some of the O2 types by having the following
in your `~/.lldbinit` (or `$PWD/.lldbinit`):

```lldb
command script import Framework/Core/scripts/lldb_o2_formatters.py
```


Before:

```gdb
(o2::framework::ConfigParamSpec &) 0x0000000774871e20: {
name = "timeframes-rate-limit-ipcid"
type = String
defaultValue = {
mStore = (__data = "\xa0\xae\x80t\a")
mType = String
mSize = 1
}
help = (str = "Suffix for IPC channel for metric-feedback, -1 = disable")
kind = kGeneric
}
```

After:

```gdb
(o2::framework::ConfigParamSpec &) 0x00000007cac75e20: {
name = "timeframes-rate-limit-ipcid"
type = String
defaultValue = {
value = 0x00000007cac0eea0 "-1"
}
help = (str = "Suffix for IPC channel for metric-feedback, -1 = disable")
kind = kGeneric
}
```
296 changes: 296 additions & 0 deletions Framework/Core/scripts/lldb_o2_formatters.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,296 @@
# Copyright 2019-2026 CERN and copyright holders of ALICE O2.
# See https://alice-o2.web.cern.ch/copyright for details of the copyright holders.
# All rights not expressly granted are reserved.
#
# This software is distributed under the terms of the GNU General Public
# License v3 (GPL Version 3), copied verbatim in the file "COPYING".
#
# In applying this license CERN does not waive the privileges and immunities
# granted to it by virtue of its status as an Intergovernmental Organization
# or submit itself to any jurisdiction.
#
# lldb data formatters for o2::framework types.
#
# Usage: add to ~/.lldbinit or a project .lldbinit:
# command script import /path/to/O2/Framework/Core/scripts/lldb_o2_formatters.py

import lldb

# o2::framework::VariantType enum values (must match Variant.h)
_VARIANT_TYPE = {
0: 'Int',
1: 'Int64',
2: 'Float',
3: 'Double',
4: 'String',
5: 'Bool',
6: 'ArrayInt',
7: 'ArrayFloat',
8: 'ArrayDouble',
9: 'ArrayBool',
10: 'ArrayString',
11: 'Array2DInt',
12: 'Array2DFloat',
13: 'Array2DDouble',
14: 'LabeledArrayInt',
15: 'LabeledArrayFloat',
16: 'LabeledArrayDouble',
17: 'UInt8',
18: 'UInt16',
19: 'UInt32',
20: 'UInt64',
21: 'Int8',
22: 'Int16',
23: 'LabeledArrayString',
24: 'Empty',
25: 'Dict',
26: 'Unknown',
}

# Map VariantType value → (C type name for FindFirstType, is_pointer_to_value)
# is_pointer_to_value=True means mStore holds a T* pointing to heap data (arrays)
_SIMPLE_TYPES = {
0: ('int', False),
1: ('long long', False),
2: ('float', False),
3: ('double', False),
5: ('bool', False),
17: ('unsigned char', False),
18: ('unsigned short',False),
19: ('unsigned int', False),
20: ('unsigned long long', False),
21: ('signed char', False),
22: ('short', False),
}

_ARRAY_ELEM_TYPES = {
6: 'int',
7: 'float',
8: 'double',
9: 'bool',
}

MAX_ARRAY_DISPLAY = 16


import struct as _struct


def _read_pointer(process, addr):
err = lldb.SBError()
ptr_size = process.GetAddressByteSize()
data = process.ReadMemory(addr, ptr_size, err)
if err.Fail() or not data:
return None
fmt = '<Q' if ptr_size == 8 else '<I'
return _struct.unpack(fmt, data)[0]


def _read_libcxx_string(process, addr):
"""Read a libc++ std::string directly from memory (no type lookup needed).

libc++ std::string on 64-bit little-endian (macOS x86_64 / arm64):
sizeof = 24 bytes, union of:
__short: byte[0] = (size<<1)|0 (short flag), bytes[1..22] = inline data
__long: byte[0] low bit = 1, bytes[8..15] = size, bytes[16..23] = data ptr
"""
err = lldb.SBError()
raw = process.ReadMemory(addr, 24, err)
if err.Fail() or not raw:
return '"<read error>"'
b = raw if isinstance(raw[0], int) else bytes(ord(c) for c in raw)
if (b[0] & 1) == 0: # short form
size = b[0] >> 1
text = b[1:1 + size].decode('utf-8', errors='replace')
else: # long form
size = _struct.unpack_from('<Q', b, 8)[0]
data_ptr = _struct.unpack_from('<Q', b, 16)[0]
if data_ptr == 0:
return '"<null>"'
heap = process.ReadMemory(data_ptr, min(size, 512), err)
if err.Fail() or not heap:
return '"<read error>"'
h = heap if isinstance(heap[0], int) else bytes(ord(c) for c in heap)
text = h[:size].decode('utf-8', errors='replace')
if size > 512:
text += '...'
return f'"{text}"'


def variant_summary(valobj, _internal_dict):
# Use GetNonSyntheticValue() so we see the real struct members even when
# the synthetic provider has replaced the children with decoded values.
raw = valobj.GetNonSyntheticValue()
mType_val = raw.GetChildMemberWithName('mType')
if not mType_val.IsValid():
return '<invalid Variant>'

mType = mType_val.GetValueAsUnsigned(26) # default Unknown
mSize = raw.GetChildMemberWithName('mSize').GetValueAsUnsigned(1)
mStore = raw.GetChildMemberWithName('mStore')
store_addr = mStore.GetLoadAddress()

type_name = _VARIANT_TYPE.get(mType, f'<type={mType}>')
target = valobj.GetTarget()
process = valobj.GetProcess()

# --- simple scalar types ---
if mType in _SIMPLE_TYPES:
ctype, _ = _SIMPLE_TYPES[mType]
t = target.FindFirstType(ctype)
if t.IsValid():
v = valobj.CreateValueFromAddress('v', store_addr, t)
return f'{type_name}({v.GetValue()})'
return f'{type_name}(?)'

# --- String (const char* stored in mStore) ---
if mType == 4:
ptr = _read_pointer(process, store_addr)
if ptr and ptr != 0:
s = _read_cstring(process, ptr)
return f'String("{s}")'
return 'String(null)'

# --- C-style numeric arrays (int*, float*, double*, bool*) ---
if mType in _ARRAY_ELEM_TYPES:
elem_type_name = _ARRAY_ELEM_TYPES[mType]
ptr = _read_pointer(process, store_addr)
if not ptr or ptr == 0:
return f'{type_name}(null)'
elem_t = target.FindFirstType(elem_type_name)
if not elem_t.IsValid():
return f'{type_name}(? x {mSize})'
count = min(mSize, MAX_ARRAY_DISPLAY)
items = []
for i in range(count):
v = valobj.CreateValueFromAddress(f'e{i}', ptr + i * elem_t.GetByteSize(), elem_t)
items.append(v.GetValue() or '?')
result = f'{type_name}([{", ".join(items)}]'
if mSize > MAX_ARRAY_DISPLAY:
result += f', ... ({mSize} total)'
result += ')'
return result

# --- ArrayString: std::vector<std::string> stored via placement new in mStore ---
if mType == 10:
# libc++ std::vector layout: __begin_, __end_, __end_cap_ (all pointers)
# libc++ std::string is always 24 bytes on 64-bit (SSO layout)
STR_SIZE = 24
ptr_size = process.GetAddressByteSize()
begin_ptr = _read_pointer(process, store_addr)
end_ptr = _read_pointer(process, store_addr + ptr_size)
if begin_ptr is None or end_ptr is None:
return 'ArrayString(?)'

count = (end_ptr - begin_ptr) // STR_SIZE if end_ptr >= begin_ptr else 0
items = []
for i in range(min(count, MAX_ARRAY_DISPLAY)):
items.append(_read_libcxx_string(process, begin_ptr + i * STR_SIZE))
result = f'ArrayString([{", ".join(items)}]'
if count > MAX_ARRAY_DISPLAY:
result += f', ... ({count} total)'
result += ')'
return result

return f'{type_name}(mSize={mSize})'


class VariantSyntheticProvider:
"""Synthetic children for o2::framework::Variant — exposes decoded value as child."""

def __init__(self, valobj, _internal_dict):
self.valobj = valobj
self.children = []

def num_children(self):
return len(self.children)

def get_child_index(self, name):
for i, (n, _) in enumerate(self.children):
if n == name:
return i
return -1

def get_child_at_index(self, index):
if 0 <= index < len(self.children):
return self.children[index][1]
return None

def update(self):
self.children = []
# Use GetNonSyntheticValue() to read the real struct members.
raw = self.valobj.GetNonSyntheticValue()
mType = raw.GetChildMemberWithName('mType').GetValueAsUnsigned(26)
mSize = raw.GetChildMemberWithName('mSize').GetValueAsUnsigned(1)
mStore = raw.GetChildMemberWithName('mStore')
store_addr = mStore.GetLoadAddress()
target = self.valobj.GetTarget()
process = self.valobj.GetProcess()

if mType in _SIMPLE_TYPES:
ctype, _ = _SIMPLE_TYPES[mType]
t = target.FindFirstType(ctype)
if t.IsValid():
v = self.valobj.CreateValueFromAddress('value', store_addr, t)
self.children.append(('value', v))

elif mType == 4: # String
ptr = _read_pointer(process, store_addr)
if ptr and ptr != 0:
char_t = target.FindFirstType('char').GetPointerType()
v = self.valobj.CreateValueFromAddress('value', store_addr, char_t)
self.children.append(('value', v))

elif mType in _ARRAY_ELEM_TYPES:
elem_type_name = _ARRAY_ELEM_TYPES[mType]
ptr = _read_pointer(process, store_addr)
if ptr and ptr != 0:
elem_t = target.FindFirstType(elem_type_name)
if elem_t.IsValid():
for i in range(min(mSize, MAX_ARRAY_DISPLAY)):
v = self.valobj.CreateValueFromAddress(f'[{i}]', ptr + i * elem_t.GetByteSize(), elem_t)
self.children.append((f'[{i}]', v))

elif mType == 10: # ArrayString
# std::vector<std::string> via placement new; std::string = 24 bytes (libc++ 64-bit)
STR_SIZE = 24
ptr_size = process.GetAddressByteSize()
begin_ptr = _read_pointer(process, store_addr)
end_ptr = _read_pointer(process, store_addr + ptr_size)
char_t = target.FindFirstType('char')
if begin_ptr is not None and end_ptr is not None and end_ptr >= begin_ptr and char_t.IsValid():
count = (end_ptr - begin_ptr) // STR_SIZE
err = lldb.SBError()
for i in range(min(count, MAX_ARRAY_DISPLAY)):
str_addr = begin_ptr + i * STR_SIZE
raw = process.ReadMemory(str_addr, STR_SIZE, err)
if err.Fail() or not raw:
continue
b = raw if isinstance(raw[0], int) else bytes(ord(c) for c in raw)
if (b[0] & 1) == 0: # short form: data inline at offset 1
data_addr = str_addr + 1
sz = max(b[0] >> 1, 1)
else: # long form: data pointer at offset 16
data_addr = _struct.unpack_from('<Q', b, 16)[0]
sz = max(_struct.unpack_from('<Q', b, 8)[0], 1)
# Expose as char[sz] so lldb renders it as a string literal
arr_t = char_t.GetArrayType(min(sz + 1, 256))
sv = self.valobj.CreateValueFromAddress(f'[{i}]', data_addr, arr_t)
self.children.append((f'[{i}]', sv))

return False # no pruning needed

def has_children(self):
return len(self.children) > 0


def __lldb_init_module(debugger, _internal_dict):
debugger.HandleCommand(
'type summary add -x "^o2::framework::Variant$" '
'--python-function lldb_o2_formatters.variant_summary'
)
debugger.HandleCommand(
'type synthetic add -x "^o2::framework::Variant$" '
'--python-class lldb_o2_formatters.VariantSyntheticProvider'
)
print('o2::framework::Variant formatters loaded.')