Fix dump-c bitfield truncation in compound assignments#8881
Open
tautschnig wants to merge 1 commit intodiffblue:developfrom
Open
Fix dump-c bitfield truncation in compound assignments#8881tautschnig wants to merge 1 commit intodiffblue:developfrom
tautschnig wants to merge 1 commit intodiffblue:developfrom
Conversation
When goto-instrument --dump-c converts a goto program back to C code,
compound assignments to bitfields (e.g., s.f &= 8 where f is a 3-bit
signed bitfield) were incorrectly handled. The goto program correctly
inserts a cast to c_bit_field to truncate the intermediate result to
the bitfield width, but goto_program2code.cpp was stripping this cast
entirely, causing the pre-truncation 32-bit value to be used in
subsequent boolean contexts.
Example: s.f = -1 (3-bit: 111), then (s.f &= 8) || 0
- Correct: -1 sign-extended & 8 = 8, truncated to 3 bits = 0, 0||0 = 0
- Bug: used pre-truncation value 8 (non-zero), so 8||0 = 1
The fix replaces the c_bit_field cast with an explicit bitwise AND mask
that truncates to the bitfield width, preserving the zero/non-zero
semantics needed for boolean contexts.
Minimal reproducer:
struct S { signed f : 3; };
static struct S s = {-1};
int main(void) { int x = (s.f &= 8) || 0; assert(x == 0); }
This was found via CSmith test seed 1684847330 (issue diffblue#7729).
Fixes: diffblue#7729
Co-authored-by: Kiro <kiro-agent@users.noreply.github.com>
tautschnig
requested review from
kroening,
martin-cs and
peterschrammel
as code owners
There was a problem hiding this comment.
Pull request overview
This PR improves --dump-c output correctness around C bit-field truncation by changing how c_bit_field typecasts are simplified during C code regeneration, and adds a regression test to catch a previously incorrect boolean-context evaluation.
Changes:
- Update
goto_program2code.cppto avoid dropping casts toc_bit_field, replacing them with explicit masking. - Add a new
goto-instrument --dump-cregression test covering compound assignment to a signed bit-field used under||.
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
src/goto-instrument/goto_program2code.cpp |
Changes cleanup_expr handling of casts to ID_c_bit_field to preserve truncation semantics in dumped C. |
regression/goto-instrument/dump-c-bitfield-truncation/test.desc |
Adds a new regression driver for --dump-c + recompile + CBMC verification. |
regression/goto-instrument/dump-c-bitfield-truncation/main.c |
Introduces a minimal reproducer using `(s.f &= 8) |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+1948
to
+1974
| const bool is_signed = bf_type.underlying_type().id() == ID_signedbv; | ||
|
|
||
| exprt op = to_typecast_expr(expr).op(); | ||
| const typet op_type = op.type(); | ||
|
|
||
| if(width == 0) | ||
| { | ||
| expr = from_integer(0, op_type); | ||
| } | ||
| else if(is_signed) | ||
| { | ||
| // For signed: mask to width bits, then sign-extend by casting | ||
| // through an unsigned type and subtracting the sign bit weight. | ||
| // Equivalent to: ((unsigned)(op) & mask) as signed, with sign | ||
| // extension. Use: (op & mask) - ((op & sign_bit) ? (1<<width) : 0) | ||
| // Simpler: just mask and let the assignment to the bitfield handle | ||
| // sign extension. But we need the VALUE to be correct for boolean | ||
| // checks. The simplest correct approach: mask the lower N bits. | ||
| // If the result is used in a boolean context, only zero/non-zero | ||
| // matters, and masking preserves that correctly. | ||
| mp_integer mask = power(mp_integer(2), mp_integer(width)) - 1; | ||
| expr = bitand_exprt(op, from_integer(mask, op_type)); | ||
| } | ||
| else | ||
| { | ||
| mp_integer mask = power(mp_integer(2), mp_integer(width)) - 1; | ||
| expr = bitand_exprt(op, from_integer(mask, op_type)); |
Comment on lines
+1959
to
+1969
| // For signed: mask to width bits, then sign-extend by casting | ||
| // through an unsigned type and subtracting the sign bit weight. | ||
| // Equivalent to: ((unsigned)(op) & mask) as signed, with sign | ||
| // extension. Use: (op & mask) - ((op & sign_bit) ? (1<<width) : 0) | ||
| // Simpler: just mask and let the assignment to the bitfield handle | ||
| // sign extension. But we need the VALUE to be correct for boolean | ||
| // checks. The simplest correct approach: mask the lower N bits. | ||
| // If the result is used in a boolean context, only zero/non-zero | ||
| // matters, and masking preserves that correctly. | ||
| mp_integer mask = power(mp_integer(2), mp_integer(width)) - 1; | ||
| expr = bitand_exprt(op, from_integer(mask, op_type)); |
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## develop #8881 +/- ##
========================================
Coverage 80.41% 80.41%
========================================
Files 1703 1703
Lines 188398 188409 +11
Branches 73 73
========================================
+ Hits 151498 151510 +12
+ Misses 36900 36899 -1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
When goto-instrument --dump-c converts a goto program back to C code, compound assignments to bitfields (e.g., s.f &= 8 where f is a 3-bit signed bitfield) were incorrectly handled. The goto program correctly inserts a cast to c_bit_field to truncate the intermediate result to the bitfield width, but goto_program2code.cpp was stripping this cast entirely, causing the pre-truncation 32-bit value to be used in subsequent boolean contexts.
Example: s.f = -1 (3-bit: 111), then (s.f &= 8) || 0
The fix replaces the c_bit_field cast with an explicit bitwise AND mask that truncates to the bitfield width, preserving the zero/non-zero semantics needed for boolean contexts.
Minimal reproducer:
struct S { signed f : 3; };
static struct S s = {-1};
int main(void) { int x = (s.f &= 8) || 0; assert(x == 0); }
This was found via CSmith test seed 1684847330 (issue #7729).
Fixes: #7729