From 36a2ee294e40a23d976c61479b2d9676ae783d98 Mon Sep 17 00:00:00 2001
From: pompolic <pompolic@special-circumstanc.es>
Date: Thu, 13 Apr 2023 16:50:46 +0200
Subject: [PATCH] Important note about stack event semantics
---
gdb-port/parser.py | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/gdb-port/parser.py b/gdb-port/parser.py
index 7db9afb..3d25e67 100644
--- a/gdb-port/parser.py
+++ b/gdb-port/parser.py
@@ -312,9 +312,9 @@ class ParserStack:
# - While a "pop" will always have a corresponding "push" event, there may be runs of "push" events
# - "pop"s will always increment self.committed, and any "push" MAY increment self.committed
# - In terms of parentheses, assuming an intact execution flow, there will never be a stray right parenthesis, but there may be an arbitrary amount of open parentheses
- # "pop" events trigger committing their allocations to the memory stats
# - For UX purposes, the left parens should also be committed as soon as possible, without double-counting the allocations made before calling h_do_parse() again, or returning
# - Working with these stack events is pretty similar to drawing flame graphs: https://brendangregg.com/FlameGraphs/memoryflamegraphs.html
+ # - semantically, the number attached to a stack event means "total allocated bytes BEFORE anything this parser does" in case of a push, whereas it means "total allocated bytes AFTER everything this parser does" with pop events
# Avoiding double-counting allocations:
# Handling the list as above will often add allocations pertaining to stack frames/parser contexts that have not finished applying yet. (All the '(' symbols at the end of the list for example)
@@ -340,6 +340,9 @@ class ParserStack:
print("commit_pop: ev_list:", ev_list) # DEBUG
print("commit_pop: parser:", parser) # DEBUG
+ for index, event in enumerate(ev_list[::-1]):
+ pass
+
def commit_pushes_at_end(self):
pass
--
GitLab