Fix Bison grammar issues preventing us from parsing CoreFoundation #5438
Conversation
hannes-steffenhagen-diffblue
requested review from
chrisr-diffblue,
kroening and
tautschnig
as code owners
Our previous attempt only worked with names like `short` or `unsigned`, but not with e.g. `unsigned short` or `uint32_t`. This patch rectifies that. Note that unlike clang we currently don't allow CV qualifiers for underlying types. Since these are ignored by clang anyway, we don't expect to see too much code in the wild making use of that feature however.
hannes-steffenhagen-diffblue
force-pushed
the
tmp/grammar-issues
branch
from
cefba9d to
ee0d3e0
Compare
| -- | ||
| gcc doesn't support postfix attributes but clang does, and Apples CoreFoundation framework makes use of this for availability macros. | ||
|
|
||
| Testing both the inline and non-inline variants because these are different cases in our parser. |
| @@ -1837,13 +1837,21 @@ enum_name: | |||
| } | |||
| ; | |||
|
|
|||
| basic_type_name_list: | |||
| basic_type_name | |||
| | basic_type_name_list basic_type_name | |||
There was a problem hiding this comment.
ℹ️ We probably want something like the following, in order to actually build the list -
| basic_type_name_list basic_type_name
{
$$=merge($1, $2);
}
;
I have added this ℹ️ comment as a note about future work which will likely be needed eventually. This is not a blocking comment.
There was a problem hiding this comment.
For onlookers: actually fully supporting this feature is in the pipeline, just getting this. to parse is higher priority though so it comes first.
hannes-steffenhagen-diffblue
force-pushed
the
tmp/grammar-issues
branch
2 times, most recently
from
c4840a3 to
712c221
Compare
With postfix attributes we mean something like
int f() __attribute__((in this position));
This is used in Apple's CoreFoundation framework so our inability to parse this
was limiting CBMC usability for code written for Apple operating systems.
hannes-steffenhagen-diffblue
force-pushed
the
tmp/grammar-issues
branch
from
712c221 to
7d2a23b
Compare
Codecov Report
@@ Coverage Diff @@
## develop #5438 +/- ##
============================================
- Coverage 68.21% 32.26% -35.96%
============================================
Files 1178 923 -255
Lines 97561 81128 -16433
============================================
- Hits 66554 26178 -40376
- Misses 31007 54950 +23943
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
Codecov Report
@@ Coverage Diff @@
## develop #5438 +/- ##
========================================
Coverage 68.21% 68.21%
========================================
Files 1178 1178
Lines 97561 97561
========================================
Hits 66554 66554
Misses 31007 31007
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
There are currently two issues preventing us from parsing Apple's CoreFoundation framework: For one, we still don't quite parse underlying enum types correctly, so for example
and
would previously fail to parse, and we didn't allow gcc-style attributes after a functions parameter list:
which gcc itself doesn't permit either but is valid in Clang's C dialect, and used this way by Apple across their frameworks for their
API_AVAILABLEandAPI_UNAVAILABLEmacros.This fixes both of these issues.