mirror of
https://github.com/Ed94/raddebugger.git
synced 2026-10-15 13:29:41 -07:00
notes
This commit is contained in:
1 parent
645cf46e55
commit
a980be7664
5 files changed
+11
-7
No files matched your search
@@ -33,7 +33,6 @@
|
||||
// [ ] need to eliminate accelerators from evaluation context, and build them on the fly instead -
|
||||
// this is necessary because, for instance, the correct locals_map varies by expression, if we
|
||||
// want to (we do) support features like "look up call stack to find local"
|
||||
// [ ] unit / module name qualification
|
||||
//
|
||||
//- memory view pass
|
||||
// [ ] toggleable ascii column
|
||||
@@ -83,8 +82,6 @@
|
||||
// a way we can defer to the underlying shell in a non-horrible way...?
|
||||
//
|
||||
//- stepping or breakpoint oddness/fixes
|
||||
// [ ] halting during a spoof-ridden step leaves the spoofs in place!!!
|
||||
// (repro via LOTS of code on one line & halting)
|
||||
// [ ] stepping-onto a line with a conditional breakpoint, which fails, causes a
|
||||
// single step over the first instruction of that line, even if the thread
|
||||
// would've stopped at the first instruction due to the step, were that bp not
|
||||
@@ -185,7 +182,6 @@
|
||||
// [ ] investigate /DEBUG:FASTLINK - can we somehow alert that we do not
|
||||
// support it?
|
||||
// [ ] just-in-time debugging
|
||||
// [ ] step-out-of-loop
|
||||
//
|
||||
//- late-conversion performance improvements
|
||||
// [ ] live++ investigations - ctrl+alt+f11 in UE?
|
||||
@@ -281,6 +277,10 @@
|
||||
// [x] unwinding in jit page - fail
|
||||
// [x] test concepts -> base
|
||||
// [x] debugger, d2r, p2r exemplar / determinism testing
|
||||
// [x] halting during a spoof-ridden step leaves the spoofs in place!!!
|
||||
// (repro via LOTS of code on one line & halting)
|
||||
// [x] unit / module name qualification
|
||||
// [x] step-out-of-loop
|
||||
|
||||
////////////////////////////////
|
||||
//~ rjf: Build Options
|
||||
|
||||
Reference in new issue
Block a user