Fix bit-field and __CPROVER_bool padding for Visual Studio #3636
Conversation
tautschnig
commented
Dec 28, 2018
tautschnig
commented
- Each commit message has a non-empty body, explaining why the change was made.
- n/a Methods or procedures I have added are documented, following the guidelines provided in CODING_STANDARD.md.
- n/a The feature or user visible behaviour I have added or modified has been documented in the User Guide in doc/cprover-manual/
- Regression or unit tests are included, or existing tests cover the modified code (in this case I have detailed which ones those are in the commit message).
- n/a My commit message includes data points confirming performance improvements (if claimed).
- My PR is restricted to a single feature or bugfix.
- n/a White-space or formatting changes outside the feature-related changed lines are in commits of their own.
tautschnig
requested review from
chrisr-diffblue,
kroening,
peterschrammel and
smowton
as code owners
tautschnig
force-pushed
the
fix-padding
branch
from
62f0be3 to
2ddac46
Compare
There was a problem hiding this comment.
✔️
Passed Diffblue compatibility checks (cbmc commit: 2ddac46).
Build URL: https://travis-ci.com/diffblue/test-gen/builds/95940092
tautschnig
force-pushed
the
fix-padding
branch
from
2ddac46 to
ed53dae
Compare
There was a problem hiding this comment.
✔️
Passed Diffblue compatibility checks (cbmc commit: ed53dae).
Build URL: https://travis-ci.com/diffblue/test-gen/builds/96063022
src/ansi-c/padding.cpp
Outdated
| @@ -201,6 +206,8 @@ static void add_padding_msvc(struct_typet &type, const namespacet &ns) | |||
| offset += (bit_field_bits + pad_bits) / config.ansi_c.char_width; | |||
| underlying_bits = bit_field_bits = 0; | |||
| } | |||
| else | |||
| offset += underlying_bits / config.ansi_c.char_width; | |||
There was a problem hiding this comment.
I think this is wrong: for example, struct { char x : 4, y : 4, z : 4, w : 4 } would have a train of 4 bitfields with underlying width 8, would get to the end of the bitfield sequence and then add 1 (underlying bits == 8 / char_width) rather than 2 to offset as it ought to. This probably ought to share the offset code with the case above, for the special case where pad_bits == 0?
There was a problem hiding this comment.
Thanks a lot, fixed and regression test based on your example added as well.
src/ansi-c/padding.cpp
Outdated
| @@ -252,6 +260,8 @@ static void add_padding_msvc(struct_typet &type, const namespacet &ns) | |||
| pad_bit_field(components, components.end(), pad); | |||
| offset += (bit_field_bits + pad) / config.ansi_c.char_width; | |||
| } | |||
| else | |||
| offset += underlying_bits / config.ansi_c.char_width; | |||
tautschnig
force-pushed
the
fix-padding
branch
from
ed53dae to
2e51ea1
Compare
There was a problem hiding this comment.
✔️
Passed Diffblue compatibility checks (cbmc commit: 2e51ea1).
Build URL: https://travis-ci.com/diffblue/test-gen/builds/96176119
tautschnig
force-pushed
the
fix-padding
branch
from
2e51ea1 to
53f0c08
Compare
tautschnig
changed the title
There was a problem hiding this comment.
✔️
Passed Diffblue compatibility checks (cbmc commit: 53f0c08).
Build URL: https://travis-ci.com/diffblue/test-gen/builds/96206271
tautschnig
force-pushed
the
fix-padding
branch
from
53f0c08 to
55a53bc
Compare
There was a problem hiding this comment.
✔️
Passed Diffblue compatibility checks (cbmc commit: 55a53bc).
Build URL: https://travis-ci.com/diffblue/test-gen/builds/96880551
tautschnig
force-pushed
the
fix-padding
branch
2 times, most recently
from
884f837 to
ae79a9a
Compare
There was a problem hiding this comment.
✔️
Passed Diffblue compatibility checks (cbmc commit: ae79a9a).
Build URL: https://travis-ci.com/diffblue/test-gen/builds/98731167
Bit fields that sum up to the underlying type need to update the offset to ensure padding suitable for the alignment is inserted.
There was a problem hiding this comment.
This PR failed Diffblue compatibility checks (cbmc commit: db0990e).
Build URL: https://travis-ci.com/diffblue/test-gen/builds/104505849
Status will be re-evaluated on next push.
Common spurious failures:
-
the cbmc commit has disappeared in the mean time (e.g. in a force-push)
-
the author is not in the list of contributors (e.g. first-time contributors).
-
the compatibility was already broken by an earlier merge.
