Skip to content

Commit

Permalink
Add patch for strict weak ordering in type comparison
Browse files Browse the repository at this point in the history
Fixes Windows Debug builds
  • Loading branch information
ekilmer committed Jan 4, 2025
1 parent 5dbc970 commit d4c2c6b
Show file tree
Hide file tree
Showing 6 changed files with 45 additions and 15 deletions.
6 changes: 3 additions & 3 deletions src/patches/HEAD/0001-Fix-UBSAN-errors-in-decompiler.patch
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
From 5b07c4118a9c6fe8f9d240419aaf5b8ce49becdf Mon Sep 17 00:00:00 2001
From 7c6e51dd1234387b98e1ad61d3f88a0565364b28 Mon Sep 17 00:00:00 2001
From: Eric Kilmer <eric.d.kilmer@gmail.com>
Date: Mon, 12 Aug 2024 12:02:35 -0400
Subject: [PATCH 1/5] Fix UBSAN errors in decompiler
Subject: [PATCH 1/6] Fix UBSAN errors in decompiler

Co-authored-by: Alex Cameron <asc@tetsuo.sh>
---
Expand Down Expand Up @@ -220,5 +220,5 @@ index 2571f55f1a..fe40e22b1b 100644
uintb true_result = ((uintb)(int32_t)f) & 0xffffffff;
uintb encoding = format.getEncoding(f);
--
2.47.0
2.47.1

Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
From 4da4a99471f73c3962b8c9d89d9a593896366a58 Mon Sep 17 00:00:00 2001
From 5e37c51ebc8a3ae0f32a3cb0049aaebafec48d7d Mon Sep 17 00:00:00 2001
From: Alex Cameron <asc@tetsuo.sh>
Date: Wed, 3 Aug 2022 20:01:18 +1000
Subject: [PATCH 2/5] Use `stroull` instead of `stroul` to parse address
Subject: [PATCH 2/6] Use `stroull` instead of `stroul` to parse address
offsets

---
Expand Down Expand Up @@ -34,5 +34,5 @@ index dbaa2e775f..72927bf379 100644
enddata = (const char *) tmpdata;
if (enddata - s.c_str() == s.size()) { // If no size or offset override
--
2.47.0
2.47.1

Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
From d2b0eaec8477ddaf7a1a971e45a6ebf8dceb57da Mon Sep 17 00:00:00 2001
From 0feb881ef1cbc9aa4639e16914b7d9ed863baadf Mon Sep 17 00:00:00 2001
From: Eric Kilmer <eric.d.kilmer@gmail.com>
Date: Tue, 29 Oct 2024 15:30:57 -0400
Subject: [PATCH 3/5] Use string resize instead of reserve
Subject: [PATCH 3/6] Use string resize instead of reserve

assign will fix up the size to hold all of what's copied
---
Expand Down Expand Up @@ -32,5 +32,5 @@ index 5f5fa0c7b3..4cd77156f2 100644
}
data.isTruncated = (numChars >= maximumChars);
--
2.47.0
2.47.1

Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
From fc506c1da2adc3fb249d7721f3c4a16c3c319056 Mon Sep 17 00:00:00 2001
From 2a8f30f31c24ecc7bd499648b97bb8b0c2705b78 Mon Sep 17 00:00:00 2001
From: Eric Kilmer <eric.d.kilmer@gmail.com>
Date: Tue, 29 Oct 2024 17:51:09 -0400
Subject: [PATCH 4/5] Ignore floating point test due to compilation differences
Subject: [PATCH 4/6] Ignore floating point test due to compilation differences

This test fails on macOS and Windows. I'm unsure whether it's an OS or
compiler issue.
Expand All @@ -24,5 +24,5 @@ index fe40e22b1b..91440e2510 100644
ASSERT_EQUALS(ff.printDecimal(f2, false), "0.33333334");
double f3 = doubleFromRawBits(0x3fd0000000000000);
--
2.47.0
2.47.1

Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
From 59254803a80df713860817b05fdc4c4487dd984f Mon Sep 17 00:00:00 2001
From 04cca72897d9088713a6e2dadb2774ad20ae7703 Mon Sep 17 00:00:00 2001
From: Eric Kilmer <eric.d.kilmer@gmail.com>
Date: Wed, 30 Oct 2024 14:26:57 -0400
Subject: [PATCH 5/5] Allow positive or negative NAN in decompiler floating
Subject: [PATCH 5/6] Allow positive or negative NAN in decompiler floating
point test

At least on Apple Silicon, this test reports positive NAN.
Expand Down Expand Up @@ -33,5 +33,5 @@ index f8108d3d32..1060a3e193 100644
<stringmatch name="Float print #14" min="1" max="1">double7 = 3.1415926535897933e-06;</stringmatch>
</decompilertest>
--
2.47.0
2.47.1

Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
From 6b57f1fc8e23c2e72afb05750d6ca11e94af9812 Mon Sep 17 00:00:00 2001
From: Eric Kilmer <eric.d.kilmer@gmail.com>
Date: Sat, 4 Jan 2025 16:50:52 -0500
Subject: [PATCH 6/6] decompiler: Fix inconsistency in type comparison operator

More specific class implementations straight return res instead of
returning the comparison to zero

This fixes an error in a Debug build on Windows for a strict weak
ordering error at runtime in the tests.
---
Ghidra/Features/Decompiler/src/decompile/cpp/type.hh | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Ghidra/Features/Decompiler/src/decompile/cpp/type.hh b/Ghidra/Features/Decompiler/src/decompile/cpp/type.hh
index 95ceaf9bb1..d677e63bee 100644
--- a/Ghidra/Features/Decompiler/src/decompile/cpp/type.hh
+++ b/Ghidra/Features/Decompiler/src/decompile/cpp/type.hh
@@ -306,7 +306,7 @@ struct DatatypeCompare {
/// Comparison operator
bool operator()(const Datatype *a,const Datatype *b) const {
int4 res = a->compareDependency(*b);
- if (res != 0) return (res<0);
+ if (res != 0) return res;
return a->getId() < b->getId(); }
};

--
2.47.1

0 comments on commit d4c2c6b

Please sign in to comment.