[Lldb-commits] [lldb] [lldb] When re-enabling a watchpoint, save the current bytes (PR #218819)

[via lldb-commits]

jimingham wrote:

> Thanks for pushing me to look more closely at the details behind how watchpoints work.
> 
> I think I want a more basic approach to refreshing the cached byte values for a Watchpoint when a user re-enables it. I created a new PR to track this simpler idea: #219333
> 
> and I am still leaning towards "lldb should delete all watchpoints on process clearing". Anyone who has a legitimately non-ASLR debug environment and Watchpoints could persist across can re-create them. Either that, or we need to push the logic for evaluating a Watchpoint variable path down into the Watchpoint object and re-calculate the addr+size each time it's user-re-enabled.
> 
> If I delete the watchpoints on process clear, even the test case from this PR isn't valuable.
> 
> What do poeple think?

It seems annoying to have to retype a complex expression or variable path every time you restart.  If we re-evaluate the expression or variable path when you re-enable a watchpoint there would be no problems with ASLR.  Presumably the user will know when that is appropriate to do.

https://github.com/llvm/llvm-project/pull/218819