Re: [Libreoffice] final and override
On 11/27/2011 08:58 PM, Noel Grandin wrote: I see the Mozilla project is adding MOZ_FINAL and MOZ_OVERRIDE macros to help control method overriding in their codebase http://whereswalden.com/2011/11/16/introducing-moz_override-to-annotate-virtual-functions-which-override-base-class-virtual-functions/ http://whereswalden.com/2011/11/26/introducing-moz_final-prevent-inheriting-from-a-class-or-prevent-overriding-a-virtual-function/ Would such things be useful additions to the LibreOffice codebase? In principle, yes. Like happened with features of the old C++ Standard that only made it slowly into the various compilers (the infamous VOS_NAMESPACE etc. stuff, e.g.), we will likely need macros for new features that can be enabled for compilers that already understand them, without breaking the code for legacy compilers. However, given that override and final will only make it into GCC 4.7 (according to http://wiki.apache.org/stdcxx/C++0xCompilerSupport), it might still be a bit early to address them. Stephan ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] final and override
On Mon, 2011-11-28 at 09:03 +0100, Stephan Bergmann wrote: VOS_NAMESPACE etc. stuff, e.g.), we will likely need macros for new features that can be enabled for compilers that already understand them, without breaking the code for legacy compilers. I believe the Mozilla guys augment this stuff by having an LLVM plugin that does the validation of the keywords on top of the compiler. That seems like a sensible approach to me. On the other hand; I'd (personally) prefer to use some defines to create the 'override' and 'final' keywords (as they will be in future). Defining them to 'virtual' or even empty would do, if they are not present in the compiler. cf. glib's provision of a stock 'inline'. That should help reduce code thrash, and hideous ugliness ;-) keeping us closer to more readable, standard C++. HTH, Michael. -- michael.me...@suse.com , Pseudo Engineer, itinerant idiot ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] final and override
On 11/28/2011 11:25 AM, Michael Meeks wrote: On the other hand; I'd (personally) prefer to use some defines to create the 'override' and 'final' keywords (as they will be in future). Defining them to 'virtual' or even empty would do, if they are not present in the compiler. cf. glib's provision of a stock 'inline'. That should help reduce code thrash, and hideous ugliness ;-) keeping us closer to more readable, standard C++. One nitpick: While it would keep the code surely more readable, it would make it less standard. For one, override and final are technically not keywords in C++11, so a correct program that used them as identifiers would be broken if we defined them to be empty (for a compiler not yet supporting them). For another, if override and final /were/ keywords, defining them in any way would result in undefined behaviour. Stephan ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] final and override
On Mon, 2011-11-28 at 11:42 +0100, Stephan Bergmann wrote: On 11/28/2011 11:25 AM, Michael Meeks wrote: On the other hand; I'd (personally) prefer to use some defines to create the 'override' and 'final' keywords (as they will be in future). Defining them to 'virtual' or even empty would do, if they are not present in the compiler. cf. glib's provision of a stock 'inline'. ... One nitpick: While it would keep the code surely more readable, it would make it less standard. For one, override and final are technically not keywords in C++11, so a correct program that used them as identifiers would be broken if we defined them to be empty (for a compiler not yet supporting them). Sure - but I suspect the instance count of that is really low in our nFinal aOverride world; a quick: git grep final | grep \.[ch]xx git grep override | grep \.[ch]xx Shows almost no problems here. For another, if override and final /were/ keywords, defining them in any way would result in undefined behaviour. Sure - which is why we'd do that for old compiler versions only; that is what glib does for 'inline' - quite successfully. The main downside to this might be the odd error messages when someone does: bool final = false; ;-) but - apparently people don't do that. Either way, I know which I prefer: finalbool methodFoo(); virtual char methodBaz(); override int methodBaa(); vs. SAL_FINALbool methodFoo() virtual char methodBaz(); SAL_OVERRIDE int methodBaa() But, since there is (as yet) no patch, it's hardly worth discussing. ;-) Michael. -- michael.me...@suse.com , Pseudo Engineer, itinerant idiot ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] final and override
On 11/28/2011 01:27 PM, Michael Meeks wrote: Sure - which is why we'd do that for old compiler versions only ...where it would still result in undefined behaviour, technically (note the nitpick, was meant somewhat tongue-in-cheek). finalbool methodFoo(); virtual char methodBaz(); override int methodBaa(); (Note that final and override go at the end of the declaration.) Stephan ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] final and override
Hello, However, given that override and final will only make it into GCC 4.7 (according to http://wiki.apache.org/stdcxx/C++0xCompilerSupport), it according to gcc.gnu.org too.. but since it seems that MSCV (according the same wiki) supports it / will support it, why not just wait a bit and then use it directly, setting a minimal compiler version ? This would allow a cleaner code, not having to take into account before implementation status. (would also be valid for other major features ) Most probably a too harsh approach ? regards Pierre-André ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [Libreoffice] final and override
On 11/28/2011 08:10 PM, Pierre-André Jacquod wrote: but since it seems that MSCV (according the same wiki) supports it / will support it, why not just wait a bit and then use it directly, setting a minimal compiler version ? This would allow a cleaner code, not having to take into account before implementation status. (would also be valid for other major features ) Most probably a too harsh approach ? We need to support rather old GCC version on Mac OS X. Stephan ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
[Libreoffice] final and override
Hi I see the Mozilla project is adding MOZ_FINAL and MOZ_OVERRIDE macros to help control method overriding in their codebase http://whereswalden.com/2011/11/16/introducing-moz_override-to-annotate-virtual-functions-which-override-base-class-virtual-functions/ http://whereswalden.com/2011/11/26/introducing-moz_final-prevent-inheriting-from-a-class-or-prevent-overriding-a-virtual-function/ Would such things be useful additions to the LibreOffice codebase? -- Noel Grandin ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
