Implement DAP setVariable request

This patch implements the DAP setVariable request.

setVariable is a bit odd in that it specifies the variable to modify
by passing in the variable's container and the name of the variable.
This approach can't handle variable shadowing (there are a couple of
open DAP bugs on this topic), so this patch renames duplicates to
avoid the problem.
This commit is contained in:
Tom Tromey
2023-10-05 06:27:50 -06:00
parent 8170af78e1
commit 87e3cc466e
6 changed files with 311 additions and 21 deletions

View File

@@ -141,3 +141,24 @@ def set_expression(
return send_gdb_with_response(
lambda: _set_expression(expression, value, frameId, format)
)
# Helper function to perform an assignment.
@in_gdb_thread
def _set_variable(ref, name, value, value_format):
with apply_format(value_format):
var = find_variable(ref)
lhs = var.find_child_by_name(name)
rhs = gdb.parse_and_eval(value)
lhs.assign(rhs)
return lhs.to_object()
@capability("supportsSetVariable")
@request("setVariable")
def set_variable(
*, variablesReference: int, name: str, value: str, format=None, **args
):
return send_gdb_with_response(
lambda: _set_variable(variablesReference, name, value, format)
)

View File

@@ -59,6 +59,9 @@ class _ScopeReference(BaseReference):
# FIXME construct a Source object
return result
def has_children(self):
return True
def child_count(self):
return len(self.var_list)

View File

@@ -16,7 +16,8 @@
import gdb
from .startup import in_gdb_thread
from .server import client_bool_capability
from abc import abstractmethod
from abc import ABC, abstractmethod
from collections import defaultdict
from contextlib import contextmanager
@@ -52,7 +53,7 @@ def apply_format(value_format):
return _null()
class BaseReference:
class BaseReference(ABC):
"""Represent a variable or a scope.
This class is just a base class, some methods must be implemented in
@@ -72,7 +73,7 @@ class BaseReference:
all_variables.append(self)
self.ref = len(all_variables)
self.name = name
self.children = None
self.reset_children()
@in_gdb_thread
def to_object(self):
@@ -80,16 +81,27 @@ class BaseReference:
The resulting object is a starting point that can be filled in
further. See the Scope or Variable types in the spec"""
result = {
"variablesReference": self.ref,
}
result = {"variablesReference": self.ref if self.has_children() else 0}
if self.name is not None:
result["name"] = str(self.name)
return result
def no_children(self):
"""Call this to declare that this variable or scope has no children."""
self.ref = 0
@abstractmethod
def has_children(self):
"""Return True if this object has children."""
return False
def reset_children(self):
"""Reset any cached information about the children of this object."""
# A list of all the children. Each child is a BaseReference
# of some kind.
self.children = None
# Map from the name of a child to a BaseReference.
self.by_name = {}
# Keep track of how many duplicates there are of a given name,
# so that unique names can be generated. Map from base name
# to a count.
self.name_counts = defaultdict(lambda: 1)
@abstractmethod
def fetch_one_child(self, index):
@@ -105,23 +117,55 @@ class BaseReference:
"""Return the number of children of this variable."""
return
# Helper method to compute the final name for a child whose base
# name is given. Updates the name_counts map. This is used to
# handle shadowing -- in DAP, the adapter is responsible for
# making sure that all the variables in a a given container have
# unique names. See
# https://github.com/microsoft/debug-adapter-protocol/issues/141
# and
# https://github.com/microsoft/debug-adapter-protocol/issues/149
def _compute_name(self, name):
if name in self.by_name:
self.name_counts[name] += 1
# In theory there's no safe way to compute a name, because
# a pretty-printer might already be generating names of
# that form. In practice I think we should not worry too
# much.
name = name + " #" + str(self.name_counts[name])
return name
@in_gdb_thread
def fetch_children(self, start, count):
"""Fetch children of this variable.
START is the starting index.
COUNT is the number to return, with 0 meaning return all."""
COUNT is the number to return, with 0 meaning return all.
Returns an iterable of some kind."""
if count == 0:
count = self.child_count()
if self.children is None:
self.children = [None] * self.child_count()
result = []
for idx in range(start, start + count):
if self.children[idx] is None:
(name, value) = self.fetch_one_child(idx)
self.children[idx] = VariableReference(name, value)
result.append(self.children[idx])
return result
name = self._compute_name(name)
var = VariableReference(name, value)
self.children[idx] = var
self.by_name[name] = var
yield self.children[idx]
@in_gdb_thread
def find_child_by_name(self, name):
"""Find a child of this variable, given its name.
Returns the value of the child, or throws if not found."""
# A lookup by name can only be done using names previously
# provided to the client, so we can simply rely on the by-name
# map here.
if name in self.by_name:
return self.by_name[name]
raise Exception("no variable named '" + name + "'")
class VariableReference(BaseReference):
@@ -135,16 +179,27 @@ class VariableReference(BaseReference):
RESULT_NAME can be used to change how the simple string result
is emitted in the result dictionary."""
super().__init__(name)
self.value = value
self.printer = gdb.printing.make_visualizer(value)
self.result_name = result_name
# We cache all the children we create.
self.value = value
self._update_value()
# Internal method to update local data when the value changes.
def _update_value(self):
self.reset_children()
self.printer = gdb.printing.make_visualizer(self.value)
self.child_cache = None
if not hasattr(self.printer, "children"):
self.no_children()
self.count = None
else:
if self.has_children():
self.count = -1
else:
self.count = None
def assign(self, value):
"""Assign VALUE to this object and update."""
self.value.assign(value)
self._update_value()
def has_children(self):
return hasattr(self.printer, "children")
def cache_children(self):
if self.child_cache is None: