New submission from STINNER Victor <vstin...@redhat.com>:
graminit.h is an header file associating Grammar entities to their values,
example:
#define stmt 269
Problem: defines symbols have no prefix and cause compilation issues on regular
C code depending where graminit.h is included. Example with ast.c:
static int
validate_stmt(stmt_ty stmt)
{ ... }
"#define stmt 269" causes a compilation issue, the preprocessor replaces the
code with:
static int
validate_stmt(stmt_ty 269)
{ ... }
... which is invalid.
Another example:
return validate_expr(exp->v.IfExp.test, Load) &&
validate_expr(exp->v.IfExp.body, Load) &&
validate_expr(exp->v.IfExp.orelse, Load);
The preprocessor replaces "exp->v.IfExp.test" with "exp->v.IfExp.305" which is
invalid...
The compile.h header file works around the issue by redefining 3 constants but
using "Py_" prefix:
/* These definitions must match corresponding definitions in graminit.h.
There's code in compile.c that checks that they are the same. */
#define Py_single_input 256
#define Py_file_input 257
#define Py_eval_input 258
For comparison, graminit.h uses:
#define single_input 256
#define file_input 257
#define eval_input 258
There are different solutions:
* Do nothing: require to include graminit.h at the right place
* Add a prefix to all symbols, ex: test => Py_test, grammar_test, etc.
* Rename problematic names like 'test' and 'stmt'
* Fix C code to avoid problematic name: 'test' is an an attribute name of a
node struct...
Note: "static const int single_input = 305;" cause a complation error on "case
single_input": "case label does not reduce to an integer constant".
IMHO adding a prefix would be a nice enhancement and a good compromise. It only
impacts the four C files which include graminit.h: future.c, ast.c, parsetok.c
and parsermodule.c.
----------
components: Interpreter Core
messages: 329517
nosy: vstinner
priority: normal
severity: normal
status: open
title: graminit.h defines very generic names like 'stmt' or 'test'
versions: Python 3.8
_______________________________________
Python tracker <rep...@bugs.python.org>
<https://bugs.python.org/issue35197>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
