URL:
<http://gna.org/patch/?2811>
Summary: Static (compile-time) assert mechanism
Project: Freeciv
Submitted by: jtn
Submitted on: Sat Jul 16 23:35:33 2011
Category: general
Priority: 5 - Normal
Status: In Progress
Privacy: Public
Assigned to: jtn
Originator Email:
Open/Closed: Open
Discussion Lock: Any
Planned Release: 2.4.0
_______________________________________________________
Details:
For maintainability, it's occasionally useful to be able to make the build
fail if certain conditions are not true, usually relating to build-time
tweakables or system limits.
Bug #18087 is a case in point; MAP_MAX_SIZE relies on MAX_DBV_LENGTH having a
certain minimum value, and this is currently commented, but it would be better
if the build failed hard when someone increased MAP_MAX_SIZE without noticing
the comment, rather than them having to debug some subtle failure. (OK, in
this case you run into a run-time assertion failure pretty quickly, but in
some other case you might not happen to tickle that code path in your
testing.)
This isn't a standard facility in the C language (yet), but there are a
number of well-known techniques for doing this in C, each with their
advantages and drawbacks (particularly generation of warnings, which matters
to us because we run with -Werror).
Googling found some surveys of available techniques here
<http://www.pixelbeat.org/programming/gcc/static_assert.html>, here
<http://www.jaggersoft.com/pubs/CVu11_3.html> (old!), and here
<http://bytes.com/topic/c/answers/758263-more-portable-compile-time-assert>,
for example.
I'm leaning toward this one:
#define FC_STATIC_ASSERT(cond, tag) enum { static_assert_ ## tag = 1 /
(!!(cond)) }
FC_STATIC_ASSERT(MAP_MAX_SIZE * 1000 <= MAX_DBV_LENGTH,
map_too_big_for_bitvector);
which, if it fails, throws the following error on my system:
In file included from ../packets.h:29,
from ../game.h:33,
from aisupport.c:26:
../map.h:571: error: division by zero
../map.h:571: error: enumerator value for
‘static_assert_map_too_big_for_bitvector’ is not an integer constant
Ugly, but effective.
(It's a shame this idiom doesn't map onto the
static_assert(constant-expression, "error message") pattern that's apparently
coming in a future C standard, but I haven't found a satisfactory technique
that does.)
_______________________________________________________
Reply to this item at:
<http://gna.org/patch/?2811>
_______________________________________________
Message sent via/by Gna!
http://gna.org/
_______________________________________________
Freeciv-dev mailing list
[email protected]
https://mail.gna.org/listinfo/freeciv-dev
