fixing Dockerfile for package versions by dubini0 · Pull Request #2 · subwire/libmatch

dubini0 · 2022-07-20T07:23:21Z

Hi! I came across your project while reading the "HALucinator" paper.
Creating docker with the original Dockerfile throws an AttributeError when I execute LibMatch.
I've found out the root cause was angr updating its attribute from graph to transition_graph.

So, I've made a new Dockerfile that installs the previous angr which has graph.

I would love to hear back from you.
Thx!

Implement optimizations identified in optimization_review.md: - Issue subwire#1: Precompute block attributes in LMD during initialization Caches graph traversal results (O(V+E)) for O(1) lookup per FunctionDiff. Falls back gracefully for LMDs created before this optimization. - Issue #9: Compute bidirectional distance matrix once _get_bidirectional_closest_matches computes the distance matrix once and extracts closest matches for both directions (A→B and B→A). - Issue subwire#3: Build symbol address hash table in LMD Converts O(n) symbol_for_addr lookup to O(1) hash table lookup. - Issue #5: Change recursion_list to recursion_set Uses set for O(1) membership test and removal instead of O(n) list ops. - Issue subwire#2: Add block_similarity() caching Caches similarity results to avoid recomputation for same block pairs. - Issue #8: Fix _is_better_match() duplicate distance computation Computes distance once and reuses for both comparisons. Expected improvement: 35-50% reduction in Phase 2 processing time when using LMDs with precomputed block attributes.

dubini0 · 2022-07-20T07:23:21Z

Hi! I came across your project while reading the "HALucinator" paper.
Creating docker with the original Dockerfile throws an AttributeError when I execute LibMatch.
I've found out the root cause was angr updating its attribute from graph to transition_graph.

So, I've made a new Dockerfile that installs the previous angr which has graph.

I would love to hear back from you.
Thx!

Implement optimizations identified in optimization_review.md: - Issue subwire#1: Precompute block attributes in LMD during initialization Caches graph traversal results (O(V+E)) for O(1) lookup per FunctionDiff. Falls back gracefully for LMDs created before this optimization. - Issue #9: Compute bidirectional distance matrix once _get_bidirectional_closest_matches computes the distance matrix once and extracts closest matches for both directions (A→B and B→A). - Issue subwire#3: Build symbol address hash table in LMD Converts O(n) symbol_for_addr lookup to O(1) hash table lookup. - Issue #5: Change recursion_list to recursion_set Uses set for O(1) membership test and removal instead of O(n) list ops. - Issue subwire#2: Add block_similarity() caching Caches similarity results to avoid recomputation for same block pairs. - Issue #8: Fix _is_better_match() duplicate distance computation Computes distance once and reuses for both comparisons. Expected improvement: 35-50% reduction in Phase 2 processing time when using LMDs with precomputed block attributes.

dubini0 · 2022-07-20T07:23:21Z

Hi! I came across your project while reading the "HALucinator" paper.
Creating docker with the original Dockerfile throws an AttributeError when I execute LibMatch.
I've found out the root cause was angr updating its attribute from graph to transition_graph.

So, I've made a new Dockerfile that installs the previous angr which has graph.

I would love to hear back from you.
Thx!

Implement optimizations identified in optimization_review.md: - Issue subwire#1: Precompute block attributes in LMD during initialization Caches graph traversal results (O(V+E)) for O(1) lookup per FunctionDiff. Falls back gracefully for LMDs created before this optimization. - Issue #9: Compute bidirectional distance matrix once _get_bidirectional_closest_matches computes the distance matrix once and extracts closest matches for both directions (A→B and B→A). - Issue subwire#3: Build symbol address hash table in LMD Converts O(n) symbol_for_addr lookup to O(1) hash table lookup. - Issue #5: Change recursion_list to recursion_set Uses set for O(1) membership test and removal instead of O(n) list ops. - Issue subwire#2: Add block_similarity() caching Caches similarity results to avoid recomputation for same block pairs. - Issue #8: Fix _is_better_match() duplicate distance computation Computes distance once and reuses for both comparisons. Expected improvement: 35-50% reduction in Phase 2 processing time when using LMDs with precomputed block attributes.

dubini0 · 2022-07-20T07:23:21Z

Hi! I came across your project while reading the "HALucinator" paper.
Creating docker with the original Dockerfile throws an AttributeError when I execute LibMatch.
I've found out the root cause was angr updating its attribute from graph to transition_graph.

So, I've made a new Dockerfile that installs the previous angr which has graph.

I would love to hear back from you.
Thx!

Implement optimizations identified in optimization_review.md: - Issue subwire#1: Precompute block attributes in LMD during initialization Caches graph traversal results (O(V+E)) for O(1) lookup per FunctionDiff. Falls back gracefully for LMDs created before this optimization. - Issue #9: Compute bidirectional distance matrix once _get_bidirectional_closest_matches computes the distance matrix once and extracts closest matches for both directions (A→B and B→A). - Issue subwire#3: Build symbol address hash table in LMD Converts O(n) symbol_for_addr lookup to O(1) hash table lookup. - Issue #5: Change recursion_list to recursion_set Uses set for O(1) membership test and removal instead of O(n) list ops. - Issue subwire#2: Add block_similarity() caching Caches similarity results to avoid recomputation for same block pairs. - Issue #8: Fix _is_better_match() duplicate distance computation Computes distance once and reuses for both comparisons. Expected improvement: 35-50% reduction in Phase 2 processing time when using LMDs with precomputed block attributes.

dubini0 · 2022-07-20T07:23:21Z

Hi! I came across your project while reading the "HALucinator" paper.
Creating docker with the original Dockerfile throws an AttributeError when I execute LibMatch.
I've found out the root cause was angr updating its attribute from graph to transition_graph.

So, I've made a new Dockerfile that installs the previous angr which has graph.

I would love to hear back from you.
Thx!

Implement optimizations identified in optimization_review.md: - Issue subwire#1: Precompute block attributes in LMD during initialization Caches graph traversal results (O(V+E)) for O(1) lookup per FunctionDiff. Falls back gracefully for LMDs created before this optimization. - Issue #9: Compute bidirectional distance matrix once _get_bidirectional_closest_matches computes the distance matrix once and extracts closest matches for both directions (A→B and B→A). - Issue subwire#3: Build symbol address hash table in LMD Converts O(n) symbol_for_addr lookup to O(1) hash table lookup. - Issue #5: Change recursion_list to recursion_set Uses set for O(1) membership test and removal instead of O(n) list ops. - Issue subwire#2: Add block_similarity() caching Caches similarity results to avoid recomputation for same block pairs. - Issue #8: Fix _is_better_match() duplicate distance computation Computes distance once and reuses for both comparisons. Expected improvement: 35-50% reduction in Phase 2 processing time when using LMDs with precomputed block attributes.

dubini0 · 2022-07-20T07:23:21Z

Hi! I came across your project while reading the "HALucinator" paper.
Creating docker with the original Dockerfile throws an AttributeError when I execute LibMatch.
I've found out the root cause was angr updating its attribute from graph to transition_graph.

So, I've made a new Dockerfile that installs the previous angr which has graph.

I would love to hear back from you.
Thx!

Implement optimizations identified in optimization_review.md: - Issue subwire#1: Precompute block attributes in LMD during initialization Caches graph traversal results (O(V+E)) for O(1) lookup per FunctionDiff. Falls back gracefully for LMDs created before this optimization. - Issue #9: Compute bidirectional distance matrix once _get_bidirectional_closest_matches computes the distance matrix once and extracts closest matches for both directions (A→B and B→A). - Issue subwire#3: Build symbol address hash table in LMD Converts O(n) symbol_for_addr lookup to O(1) hash table lookup. - Issue #5: Change recursion_list to recursion_set Uses set for O(1) membership test and removal instead of O(n) list ops. - Issue subwire#2: Add block_similarity() caching Caches similarity results to avoid recomputation for same block pairs. - Issue #8: Fix _is_better_match() duplicate distance computation Computes distance once and reuses for both comparisons. Expected improvement: 35-50% reduction in Phase 2 processing time when using LMDs with precomputed block attributes.

dubini0 · 2022-07-20T07:23:21Z

Hi! I came across your project while reading the "HALucinator" paper.
Creating docker with the original Dockerfile throws an AttributeError when I execute LibMatch.
I've found out the root cause was angr updating its attribute from graph to transition_graph.

So, I've made a new Dockerfile that installs the previous angr which has graph.

I would love to hear back from you.
Thx!

Implement optimizations identified in optimization_review.md: - Issue subwire#1: Precompute block attributes in LMD during initialization Caches graph traversal results (O(V+E)) for O(1) lookup per FunctionDiff. Falls back gracefully for LMDs created before this optimization. - Issue #9: Compute bidirectional distance matrix once _get_bidirectional_closest_matches computes the distance matrix once and extracts closest matches for both directions (A→B and B→A). - Issue subwire#3: Build symbol address hash table in LMD Converts O(n) symbol_for_addr lookup to O(1) hash table lookup. - Issue #5: Change recursion_list to recursion_set Uses set for O(1) membership test and removal instead of O(n) list ops. - Issue subwire#2: Add block_similarity() caching Caches similarity results to avoid recomputation for same block pairs. - Issue #8: Fix _is_better_match() duplicate distance computation Computes distance once and reuses for both comparisons. Expected improvement: 35-50% reduction in Phase 2 processing time when using LMDs with precomputed block attributes.

dubini0 · 2022-07-20T07:23:21Z

Hi! I came across your project while reading the "HALucinator" paper.
Creating docker with the original Dockerfile throws an AttributeError when I execute LibMatch.
I've found out the root cause was angr updating its attribute from graph to transition_graph.

So, I've made a new Dockerfile that installs the previous angr which has graph.

I would love to hear back from you.
Thx!

Implement optimizations identified in optimization_review.md: - Issue subwire#1: Precompute block attributes in LMD during initialization Caches graph traversal results (O(V+E)) for O(1) lookup per FunctionDiff. Falls back gracefully for LMDs created before this optimization. - Issue #9: Compute bidirectional distance matrix once _get_bidirectional_closest_matches computes the distance matrix once and extracts closest matches for both directions (A→B and B→A). - Issue subwire#3: Build symbol address hash table in LMD Converts O(n) symbol_for_addr lookup to O(1) hash table lookup. - Issue #5: Change recursion_list to recursion_set Uses set for O(1) membership test and removal instead of O(n) list ops. - Issue subwire#2: Add block_similarity() caching Caches similarity results to avoid recomputation for same block pairs. - Issue #8: Fix _is_better_match() duplicate distance computation Computes distance once and reuses for both comparisons. Expected improvement: 35-50% reduction in Phase 2 processing time when using LMDs with precomputed block attributes.

Update Dockerfile
89afcfd