jimingham wrote:
I think this patch is fine but I have some quibbles about the description of
the feature in the comments.
In the context of ValueObjectVariable, it's appropriate to speak about what
DWARF tells you about the nature of the variable, and how that would inform the
decision about whether the value can be altered or not.
But you use the same language in other ValueObject classes where it isn't as
appropriate. For instance I could make a value object wholly out of a
DataBufferHeap and I would expect to be able to change that ValueObject even
though it has no representation in the target. So requiring memory isn't a
necessary condition for variables being editable.
As another example, ValueObjectConstResult ValueObjects often mirror the value
that was stored on the host side to a value in the target. For the ones we
make for "result variables" we artificially mark them as constant (with
SetIsConstant). But the ones we make for expression result variables are
modifiable. For instance, with:
```
2 int
3 print_ref(int *int_ptr) {
4 *int_ptr += 11;
5 return printf("I got: %d\n", *int_ptr);
6 }
```
I can do:
```
(lldb) expr int $my_int = 10
(lldb) expr print_ref(&$my_int)
I got: 21
(int) $0 = 10
(lldb) expr $my_int
(int) $my_int = 21
(lldb) expr $my_int = 30
(int) $1 = 30
(lldb) expr print_ref(&$my_int)
I got: 41
```
So we were able to pass the value by address to a function we actually called,
and change its value and effect the value in the target. This doesn't happen
for expression result variables because we don't copy back the results of
expressions. We really should also reset the value, but we don't, leading to:
```
(lldb) expr argc
(int) $3 = 1
(lldb) expr print_ref(&$3)
I got: 12
(int) $4 = 10
(lldb) expr $3
(int) $3 = 1
(lldb) expr print_ref(&$3)
I got: 23
(int) $5 = 10
```
The fact that we said $3 was still 1 but the expression that used the
target-side representation saw 12 is a bug.
That was way too many words for the suggestion that the discussion of what
makes ValueObjectVariables not changeable is not what makes other ValueObject
subclasses changeable...
https://github.com/llvm/llvm-project/pull/208042
_______________________________________________
lldb-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
