C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Sutters guidelines and redundant include guards          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message apm35@student.open.ac.uk Guest Posted: Tue Dec 14, 2004 7:54 pm    Post subject: Sutters guidelines and redundant include guards I have just come across the guideline: "Always write internal #include guards. Never write external #include guards." and I have a few comments about it: 1. Sutter's example is not quite correct. The code fragment he gives is for headers that do not have a predictable guard (e.g system headers or third-party headers). When the guard is predictable the construct is (slightly) smaller, e.g: #ifndef INCLUDED_FOOBAR_H #include "foobar.h" #endif 2. Microsoft compilers and other compilers aimed at Windoze machines do indeed make redundant include guards, er, redundant. However, Unix compilers have a long way to go before they catch up in this area. There maybe perhaps could still be some benefit from using them for code that has to compile on Unix. 3. IMO the predictable aspect of the include guard has been glossed over, leaving the impression that the redundant guard construct is high-maintainance. Well it would be if the guard was not predictable. But it is predictable, using whatever rules are established in the corporate coding stds. This reduces the maintainance costs and leaves one with what is probably the main objection: the code is longer and more ugly. This is still an understandable reason for avoiding them but is less strong than Sutter's guideline suggests. 4. Given the above, whether to use them or not boils down to whether or not the improved builds times is a gain that is worth the cost of the code uglification. This would have to be measured. Given the predictable mechanical nature of the redundant guards surely there must be a tool by now that would convert to use redundant guards, then we could measure the improvement. So the guideline seems to come down a little heavy on a measure that could have some benefit. Regards, Andrew Marlow [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Dave O'Hearn Guest Posted: Wed Dec 15, 2004 4:44 am    Post subject: Re: Sutters guidelines and redundant include guards [email]apm35 (AT) student (DOT) open.ac.uk[/email] wrote: Quote: 1. Sutter's example is not quite correct. The code fragment he gives is for headers that do not have a predictable guard (e.g system headers or third-party headers). When the guard is predictable the construct is (slightly) smaller, e.g: #ifndef INCLUDED_FOOBAR_H #include "foobar.h" #endif Why would you need to write the external guard, though? Maybe you weren't aware of the optimizations done by UNIX compilers? Quote: 2. Microsoft compilers and other compilers aimed at Windoze machines do indeed make redundant include guards, er, redundant. However, Unix compilers have a long way to go before they catch up in this area. There maybe perhaps could still be some benefit from using them for code that has to compile on Unix. Unfortunately, #pragma once is not standard, and it's not just an optimization. Often, including multiple times generates wrong code, like violating the ODR. You still need the redundant include guards for portable code. Most UNIX compilers have their own optimization, instead of #pragma once, and it is better. They recognize the structure of a redundant include guard, and will not read the file multiple times, as they understand it's a no-op. This fails gracefully on compilers that don't support the optimization. However, Windows compilers tend not to support the optimization, so there is a lot of code like this, #ifndef SOMEFILE_H #define SOMEFILE_H #ifndef LACKS_PRAGMA_ONCE #pragma once #endif // !LACKS_PRAGMA_ONCE // ... #endif // !SOMEFILE_H I believe ACE starts every hearder with a construct like that. It does work though, and it should eliminate any motive for external include guards unless you're using a terribly dated compiler. -- Dave O'Hearn [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Ben Hutchings Guest Posted: Wed Dec 15, 2004 4:45 am    Post subject: Re: Sutters guidelines and redundant include guards [email]apm35 (AT) student (DOT) open.ac.uk[/email] wrote: Quote: I have just come across the guideline: "Always write internal #include guards. Never write external #include guards." and I have a few comments about it: 1. Sutter's example is not quite correct. The code fragment he gives is for headers that do not have a predictable guard (e.g system headers or third-party headers). When the guard is predictable the construct is (slightly) smaller, e.g: #ifndef INCLUDED_FOOBAR_H #include "foobar.h" #endif 2. Microsoft compilers and other compilers aimed at Windoze machines do indeed make redundant include guards, er, redundant. Assuming you don't include the a header file through two different paths that it doesn't realise are the same - though that would be rather silly. Quote: However, Unix compilers have a long way to go before they catch up in this area. snip GCC's preprocessor recognises the ifndef...#define...#endif pattern #and can skip the whole file if it's included again. Any other implementer is free to do the same. There's no need for a non-standard extension, or the external guards. This also avoids the problem of having to know for certain whether two paths point to the same file. -- Ben Hutchings compatible: Gracefully accepts erroneous data from any source [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Allan W Guest Posted: Wed Dec 15, 2004 6:16 pm    Post subject: Re: Sutters guidelines and redundant include guards I know that not every platform supports #pragma once. But does anybody know of a platform where it does something undesirable? IMO, warnings would count as undesirable if there was no easy way to supress them. But according to the standard (16.6): Any pragma that is not recognized by the implementation is ignored. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top kanze@gabi-soft.fr Guest Posted: Wed Dec 15, 2004 6:17 pm    Post subject: Re: Sutters guidelines and redundant include guards Ben Hutchings wrote: Quote: GCC's preprocessor recognises the ifndef...#define...#endif pattern #and can skip the whole file if it's included again. Any other implementer is free to do the same. There's no need for a non-standard extension, or the external guards. This also avoids the problem of having to know for certain whether two paths point to the same file. It's worth noting that GCC has done this for quite some time now, at least ten years. It's surprising that so few other compilers have adopted the technique. -- James Kanze GABI Software http://www.gabi-soft.fr Conseils en informatique orientée objet/ Beratung in objektorientierter Datenverarbeitung 9 place Sémard, 78210 St.-Cyr-l'École, France, +33 (0)1 30 23 00 34 [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Dave O'Hearn Guest Posted: Thu Dec 16, 2004 1:20 pm    Post subject: Re: Sutters guidelines and redundant include guards Allan W wrote: Quote: I know that not every platform supports #pragma once. But does anybody know of a platform where it does something undesirable? IMO, warnings would count as undesirable if there was no easy way to supress them. But according to the standard (16.6): Any pragma that is not recognized by the implementation is ignored. I put it in the !LACKS_PRAGMA_ONCE test because older versions of gcc issue a warning that "`#pragma once' is obsolete". It's annoying, and I don't know how to disable it. gcc 3.4 has correct support for pragma once and doesn't issue the warning. Both gcc and MSVC++ issue warnings on unrecognized pragmas, but they both have ways to disable them. In general though, I just find pragmas scary, and I like to #ifdef them. -- Dave O'Hearn [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top apm35@student.open.ac.uk Guest Posted: Thu Dec 16, 2004 1:32 pm    Post subject: Re: Sutters guidelines and redundant include guards Dave O'Hearn wrote: Quote: including multiple times generates wrong code, like violating the ODR. You still need the redundant include guards for portable code. I think you are confusing redundant include guards with conventional include guards. Conventional include guards are used to protect againts harmful results if a file is included multiple times in a TU. This is done by the file beginning with #ifndef guardname #define guardname blah-blah #endif. This protects against ODR etc. Redundant include guards are an ADDITIONAL measure where the include guard macro is tested to ensure that the #include is not performed if the inclusion has already been made. With conventional include guards the header may be parsed many times but each time other than the first is effectively a no-op. However, the use of conventional include guards without the additional measure of redundant include guards means that extra work is done in finding and opening the file and parsing its contents (albeit the contents will be quickly scanned until the closing #endif is encountered). Quote: Most UNIX compilers have their own optimization, instead of #pragma once, and it is better. They recognize the structure of a redundant include guard, They recognise the construct of a *conventional* include guard. And so the cost of finding and opening the file (which is large compared to the cost of skipping to the closing #endif) is still incurred. Quote: work though, and it should eliminate any motive for external include guards unless you're using a terribly dated compiler. One of my points was that commercial unix compilers are not exactly leading edge and so do not (AFAIK) have any work done in this area. -Andrew M. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Ben Hutchings Guest Posted: Fri Dec 17, 2004 4:40 am    Post subject: Re: Sutters guidelines and redundant include guards [email]apm35 (AT) student (DOT) open.ac.uk[/email] wrote: Quote: Dave O'Hearn wrote: including multiple times generates wrong code, like violating the ODR. You still need the redundant include guards for portable code. snip With conventional include guards the header may be parsed many times but each time other than the first is effectively a no-op. However, the use of conventional include guards without the additional measure of redundant include guards means that extra work is done in finding and opening the file and parsing its contents (albeit the contents will be quickly scanned until the closing #endif is encountered). snip A smart preprocessor can identify the guard macro the first time a file is included and check it the second time before opening the file. (It is reasonable to assume that source files are not modified during compilation.) This is what GCC's preprocessor does. See <http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/libcpp/files.c?rev=1.1.2.3> and in particular the function should_stack_file. -- Ben Hutchings This sentence contradicts itself - no actually it doesn't. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First         C++Talk.NET Forum Index -> C++ Language (Moderated) All times are GMT Page 1 of 1   Jump to: Select a forum C/C++----------------C LanguageC Language (Moderated)C++ Language (Moderated)C++ language (comp.lang.c++)C++ language, library and standards Non-English C/C++ Newsgroups----------------C++ (German)C++ (French)  You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum Powered by phpBB © 2001, 2006 phpBB Group SEO toolkit © 2004-2006 webmedic.