jasonmolenda wrote:
> > Does `watch enable` not re-evaluate the variable expression that was passed
> > in when the watchpoint was set? I was wondering why you have to require
> > ASLR?
>
> _cough_ I forgot how all this works, I had to reread a bit. We don't
> reevaluate the watchpoint specification when re-enabling. In the Watchpoint
> object the specification is resolved to an address and a size, and the
> WatchpointResource(s) for that Watchpoint are based on that addr+size.
Ah, Watchpoint itself never interprets the specification, at least not watch
set variable. That's done in CommandObjectWatchpointSetVariable::DoExecute,
```
valobj_sp = frame->GetValueForVariableExpressionPath(
command.GetArgumentAtIndex(0), eNoDynamicValues, expr_path_options,
var_sp, error);
auto [addr, addr_type] = valobj_sp->GetAddressOf(false);
WatchpointSP watch_sp =
target->CreateWatchpoint(addr, size, &compiler_type, watch_type, error);
if (var_sp->GetScope() == eValueTypeVariableLocal)
watch_sp->SetupVariableWatchpointDisabler(m_exe_ctx.GetFrameSP());
```
So re-enabling a watchpoint will never re-evaluate the expression used to
create it. Maybe it should, but I'd probably redesign the Watchpoint ctor from
"requires an addr & size" to optionally only giving a specification string to
be evaluated in an ExecutionContext, and let Watchpoint itself resolve this to
an addr+size, and re-do that when re-enabling (with the new ExecutionContext
of course).
https://github.com/llvm/llvm-project/pull/218819
_______________________________________________
lldb-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
