[Lldb-commits] [lldb] [lldb] [mostly NFC] Large WP foundation: WatchpointResources (PR #68845)

[Jason Molenda via lldb-commits]

================
@@ -2,8 +2,8 @@
 #include <stdio.h>
 
 int main() {
-  uint8_t x1 = 0;
-  uint16_t x2 = 0;
+  long x1 = 0;
+  long x2 = 0;
----------------
jasonmolenda wrote:

You're right.  I thought this test was *unintentionally* testing the case where we have two hardware watchpoint register watching the same doubleword of memory.  But that was actually the point of the test (v. https://reviews.llvm.org/D77173 ).  I'm not sure how thrilled I am about this test, it seems to be encoding the behavior of the cpu with multiple watchpoint registers on the same doubleword and I don't know how universal that is - if a CPU might only evaluate the first watchpoint on a doubleword and ignore the second.  

This is very much in the wheelhouse of the work I need to do on large watchpoints where a single WatchpointResource should be used for both watchpoints, expanded to include the byte range of both, and then reduced back down again when one is disabled or deleted.  I'll be adding tests like this one as part of that work.  I'll revert my change here.

https://github.com/llvm/llvm-project/pull/68845