Re: remove server silliniess
On April 16, 2003 01:51 pm, Gregory M. Turner wrote:
remove run-time checks of compile-time constants.
Why is that a silliness? I dearly wish that there were a way of doing compile-time assertion, but there isn't, so it has to be done by a runtime assertion. So it adds a tiny amount to the runtime. If that really bugs you then create a function of compile-time assertions and ensure that it gets called once only. Personally considering the current status of wine and the way we have to work on code that we don't really understand I am surprised at how few assertions there are (and lets not get started on internal code documentation) Or are you saying that assertions are a waste of time :-) -- Bill Medland ACCPAC International, Inc. medbi01(a)accpac.com Corporate: www.accpac.com Hosted Services: www.accpaconline.com
Bill Medland <billmedland(a)mercuryspeed.com> writes:
So it adds a tiny amount to the runtime. If that really bugs you then create a function of compile-time assertions and ensure that it gets called once only.
Actually it doesn't add anything at all to the runtime, it gets optimized out. -- Alexandre Julliard julliard(a)winehq.com
On April 16, 2003 04:30 pm, Alexandre Julliard wrote:
Bill Medland <billmedland(a)mercuryspeed.com> writes:
So it adds a tiny amount to the runtime. If that really bugs you then create a function of compile-time assertions and ensure that it gets called once only.
Actually it doesn't add anything at all to the runtime, it gets optimized out.
sheepish grin!! I never thought of that. We learn something new every day. -- Bill Medland ACCPAC International, Inc. medbi01(a)accpac.com Corporate: www.accpac.com Hosted Services: www.accpaconline.com
remove run-time checks of compile-time constants.
I dearly wish that there were a way of doing compile-time assertion, but there isn't, so it has to be done by a runtime assertion.
Well, there seem to be some reasonable tricks for that: http://www.jaggersoft.com/pubs/CVu11_3.html http://www.panelsoft.com/murphyslaw/apr01.htm This is for C++ but maybe also adaptable for C: http://www.boost.org/libs/static_assert/static_assert.htm bye Fabi
On Thursday 17 April 2003 02:13 am, Fabian Cenedese wrote:
remove run-time checks of compile-time constants.
I dearly wish that there were a way of doing compile-time assertion, but there isn't, so it has to be done by a runtime assertion.
Well, there seem to be some reasonable tricks for that: http://www.jaggersoft.com/pubs/CVu11_3.html http://www.panelsoft.com/murphyslaw/apr01.htm
This is for C++ but maybe also adaptable for C: http://www.boost.org/libs/static_assert/static_assert.htm
bye Fabi
ah, the hidden powers of the preprocessor -- how it drives us to obsfucation with it's robotic stupidity... I wonder, is cpp a turing machine? Makes me want to implement vim as a sed script ;) nonetheless, considering one must sometimes make do with what one has, that array-of-zero trick is super-clever... perhaps I'll actually make a constructive patch out of this debacle yet.. -- gmt "Democracy is the theory that holds that the common people know what they want, and deserve to get it good and hard." HL Mencken
Well, there seem to be some reasonable tricks for that: http://www.jaggersoft.com/pubs/CVu11_3.html http://www.panelsoft.com/murphyslaw/apr01.htm
This is for C++ but maybe also adaptable for C: http://www.boost.org/libs/static_assert/static_assert.htm
Here is my version, which has some advantages over the mentioned compile time asserts. It uses the constraint, that a typedef'd array can't have zero length. -> It doesn't generate any code in the compiled binary. (no use of switch or other constructs) -> It doesn't use any memory in the running program. (no declaration of variables) It generates a unique name for the typedef. -> You can use any number of assertions, if you put any other in its own source code line. You don't have to put the assert into a function. You can put it anywhere at global scope #define MAKE_UNIQUE_(x,y) __unique_name_##x##y #define MAKE_UNIQUE(x,y) MAKE_UNIQUE_(x,y) #define UNIQUE_NAME(name) MAKE_UNIQUE(name, __LINE__) /* compile time assertions; An assertion failure results in error C2466: cannot allocate an array of constant size 0 */ #define CASSERT(expr) typedef char UNIQUE_NAME(cassert_type)[(expr)? 1: 0]; /* Examples */ CASSERT(sizeof(char)>=1); CASSERT(sizeof(int)>=2); CASSERT(sizeof(long)=4); CASSERT(sizeof(__int64)=8); Using the CASSERT macro you can define some more convenient macros for checking type sizes: #include <limits.h> #define ASSERT_MIN_BITSIZE(type, size) CASSERT(sizeof(type)*CHAR_BIT >= (size)); #define ASSERT_EXACT_BITSIZE(type, size) CASSERT(sizeof(type)*CHAR_BIT = (size)); /* Examples */ ASSERT_MIN_BITSIZE(char, 8); ASSERT_MIN_BITSIZE(int, 16); ASSERT_EXACT_BITSIZE(long, 32); ASSERT_EXACT_BITSIZE(__int64, 64); -- Martin Fuchs martin-fuchs(a)gmx.net
participants (5)
-
Alexandre Julliard -
Bill Medland -
Fabian Cenedese -
Gregory M. Turner -
Martin Fuchs