- copying an aggregate (array) - 3 Updates
- copying an aggregate (array) - 1 Update
| James Kuyper <jameskuyper@verizon.net>: Dec 02 03:18PM -0500 On 12/01/2017 10:45 PM, Stefan Ram wrote: > the C++ specification where this is allowed, but I also think > that the general idea is that there is a general license to > put names into namespaces and a restriction regarding »::std«. I too take it for granted, I ask only because it has been repeatedly asserted that ::std is pointless and unnecessary, an assertion which would only make sense to me if std were a reserved identifier for use as an namespace name. Basically, the response that I've gotten is that std is, for practical purposes, a reserved identifier - not because code which violates such a reservation is a constraint violation, but because delivering such code is likely to get you fired. While I can appreciate that argument, it's not directly relevant to my point. If it is indeed such a bad idea to do that, I think it should be identified as a constraint violation. So long as it isn't a constraint violation, fully conforming implementations must accept and correctly translate code which defines such a namespace, and I get the impression from some of the things that have been said, that many implementation of the C++ standard library might malfunction if used with code that defined such a namespace. |
| Chris Vine <chris@cvine--nospam--.freeserve.co.uk>: Dec 02 09:56PM On Sat, 2 Dec 2017 15:18:47 -0500 > that have been said, that many implementation of the C++ standard > library might malfunction if used with code that defined such a > namespace. The C++ standard cannot explicitly rule out every case of bad programming practice, nor for that matter every possible case of malicious coding by a psychopathic programmer, if that is what you mean by "constraint violation". How long do you want the standard to be? Only an incompetent or malicious programmer would place a 'std' namespace within another namespace given current coding practices, which treat 'std' as being at the top level. If a project is so concerned about this it could put that in its coding standard for the project, but the idea of using 'std' in this way is so outlandish that I would view it as unnecessary. Writing code in any language requires programmers to behave reasonably and in accordance with coding standards for the project in question. If they try to sabotage the project you fire them. That seems perfectly reasonable to me. Chris |
| James Kuyper <jameskuyper@verizon.net>: Dec 02 06:15PM -0500 On 12/02/2017 04:56 PM, Chris Vine wrote: >>> "James R. Kuyper" <jameskuyper@verizon.net> writes: >>>> believe it says nothing about, for example, ::kuyper::std. if I'm >>>> right about that, that's sufficient to allow such a namespace. ... > programming practice, nor for that matter every possible case of > malicious coding by a psychopathic programmer, if that is what you mean > by "constraint violation". I tend to spend most of my time on the C newsgroups; I forgot that "constraint violation" is a C-specific term. The key feature of a constraint that matters in this context is that at least one diagnostic must be issued for any program that contains any syntax errors or constraint violations, and that either of those features means that is an implementation is no longer required to accept the program. The C++ standard imposes the same requirements for any violation of any diagnosable rule (1.4p2). The C standard distinguishes constraints from other kinds of rules; the C++ standard makes the same distinction by specifying, for a large number of different rules, that "no diagnostic is required" - which seems to me to be a clumsier way of making the same distinction. Creation of such a namespace is trivially diagnosable, so I'm just suggesting that a rule be created prohibiting it - if such a rule accurately represents what the committee believes the standard should say. > How long do you want the standard to be? That's the wrong question. The key point is what a conforming implementation is required to do. If you believe that a conforming implementation of the C++ standard library is allowed to use std:: rather than ::std::, even in contexts where such code could break if I defined a ::kuyper::std namespace, then the standard should explicitly state, in some fashion, that there's a problem with me defining such a namespace. If an implementation of the C++ standard library should be required to use ::std:: whenever my creation of such a namespace might otherwise cause something to break, the current wording is sufficient. > Only an incompetent or malicious programmer would place a 'std' The standard as currently written implies that every guarantee the standard makes about the behavior of my program will still hold, even if I choose to define ::kuyper::std. If my own code doesn't contain any constructs for which definition of such a namespace is problematic, that should be sufficient to have a well-formed program, regardless of how poorly you might think of me for writing it. This implies that any part of the C++ standard library that I choose to use in my program must also avoid any use of constructs that would be broken by my definition of ::kuyper::std. > concerned about this it could put that in its coding standard for the > project, but the idea of using 'std' in this way is so outlandish that > I would view it as unnecessary. I really don't care how badly you might think of me for writing such code, and I almost certainly wouldn't write such code. But I still believe that the standard should be clear about whether or not the C++ standard library is required to be written in such a way that it would not be broken by such code. > Writing code in any language requires programmers to behave reasonably > and in accordance with coding standards for the project in question. Actually, no it doesn't. Nothing requires that my program be part of a project with non-trivial coding standards. All that is actually required is that the code be, in C++ terms, well-formed, in order to ensure that it must have the behavior that the relevant standard specifies for such code. I don't have to be reasonable, and I don't have to conform to anyone else's coding standards, in order for that requirement to apply. |
| ram@zedat.fu-berlin.de (Stefan Ram): Dec 02 03:45AM >believe it says nothing about, for example, ::kuyper::std. if I'm right >about that, that's sufficient to allow such a namespace. I believe that this has been discussed in the Usenet before (in this group?) and the common wisdom was then that one may not put names into »::std«, but into any other namespace »std«. I took this for granted for a long time and am surprised that you now ask about this. I can't put my finger on a place in the C++ specification where this is allowed, but I also think that the general idea is that there is a general license to put names into namespaces and a restriction regarding »::std«. |
| You received this digest because you're subscribed to updates for this group. You can change your settings on the group membership page. To unsubscribe from this group and stop receiving emails from it send an email to comp.lang.c+++unsubscribe@googlegroups.com. |
No comments:
Post a Comment