Re:How to abuse a C++ compiler?
by vinkelhake on Thursday March 20, 2014 @10:27AM
(#7435691)
It should be noted that at least some cases which previously required template meta programming can now be done in a more natural style with constexpr. A constexpr function can be evaluated at compile time, this means that its return value can be used in places that require a compile time constant.
constexpr functions where quite restrictive in C++11. They could pretty much only contain a single expression and a return. C++14 significantly relaxes these restrictions.
it calls a function at compile time to expand placeholders, and then compiles the generated string via the mixin() statement and carries on with the code. far more pleasant than c++.
Re:How to abuse a C++ compiler?
by pcwalton on Thursday March 20, 2014 @11:37AM
(#7436185)
constexpr doesn't really help for logic programming though, which is what this post is using template specialization to do. You can't ask for the answer to Russell's paradox using constexpr (which is, IMHO, a good thing).
Re:How to abuse a C++ compiler?
by xedarius on Thursday March 20, 2014 @09:43AM
(#7435450)
I'd say C++ templates are a clear winner in this contest. Simple due to the sheer volume of unintelligible nonsense the compiler produces if there is an error in the template.
Re:How to abuse a C++ compiler?
by marcosdumay on Thursday March 20, 2014 @12:52PM
(#7436806)
That's because it is awkward.
One of the fundamental properties people expect from a step called "compilation" is that it finishes. Also templates notation wasn't designed for this kind of use, and does not scale well for complex programs.
In Lisp code generation is done in Lisp, a notation that scales quite well, and there isn't a separarted step called "compilation". There is no general assumption that "execution" finishes.
Re:How to abuse a C++ compiler?
by knome on Thursday March 20, 2014 @09:49AM
(#7435481)
It looks like clang creates it's data structure representing the template instantiation, caches the instantiation for the parameterizing types, then begins instantiating the types for the members. When it first hits the static bool, it reserves space for it's static value, then recurses into the static bools type in order to determine the value. Since the recursion uses the same parameterizing types, it finds the type it was midway through constructing, finds the space-reserved-but-no-appropriate-value-set static member in it, takes a 0 from that memory ( presumably from having allocated and zeroed the memory when initially saving space for the final value ).
For the MutualRec it may not set the static const value as being derived from a constexpr until after having retrieved the value, causing it to access the member in the cache before it's constexpr'ness has been determined.
For the Contra example, it would return 1 from having not'ed the 0 it found in the uninitialized static const.
Russel's paradox would be subverted in the same manner as the Contra example, returning true because it found false ( 0 ) in an uninitialized member variable after looking up the partially instatiated template instantiation in a cache.
This is all conjecture, but it fits, and it's where I would look first if debugging this. I do not know whether this is allowed by the standard, but I'd wager they did not account for referring cyclically to partially instantiated template parameterizations in a cache. GCC's errorful death is probably the correct action here, rather than clang's return of spurious values.
Re:How to abuse a C++ compiler?
by pbsd on Thursday March 20, 2014 @10:10AM
(#7435584)
You are correct. The Contra example is not different from the much simpler program
#include <iostream>
int main()
{
static bool p = !p;
std::cout << p << std::endl;
}
The static storage specification ensures that p is initialized to 0 before main is executed, and the standard's scoping rules (§3.3.2) allows a variable to reference itself right after declaration.
The same thing happens with Russell's Paradox, with the added twist that HasElement<M, M> always resolves to the specialization HasElement<M, T> (it 'fits' better) rather than the general version, which is never used.
Re:How to abuse a C++ compiler?
by knome on Thursday March 20, 2014 @11:49AM
(#7436265)
~/test/O$ cat break.cpp
#include <iostream>
template <typename A, typename B>
struct MutualRecursion {
static const int a = MutualRecursion< B, A >::b ;
static const int b = MutualRecursion< B, A >::a ;
};
int main( int argc, char ** argv ){
std::cout << MutualRecursion< float, int >::b << std::endl;
return 0;
};
~/test/O$ clang break.cpp
break.cpp:6:49: error: no member named 'b' in 'MutualRecursion<float, int>'
static const int a = MutualRecursion< B, A >::b ;
~~~~~~~~~~~~~~~~~~~~~~~~~^
break.cpp:6:24: note: in instantiation of template class 'MutualRecursion<int, float>' requested here
static const int a = MutualRecursion< B, A >::b ;
^
break.cpp:11:16: note: in instantiation of template class 'MutualRecursion<float, int>' requested here
std::cout << MutualRecursion< float, int >::b << std::endl;
^
break.cpp:7:24: error: in-class initializer is not a constant expression
static const int b = MutualRecursion< B, A >::a ;
^~~~~~~~~~~~~~~~~~~~~~~~~~
2 errors generated.
~/test/O$
~~This is definitely a bug in clang~~. clang doesn't find the member "b" here because it's referencing the cached still-in-construction version of the template. Furthermore, it still does this even when "b" is a regular constant.
// static const int b = MutualRecursion< B, A >::a ;
static const int b = 10 ;
gcc, since it appears to always instantiate a new copy of the template compiles successfully for when "b = 10".
~~clang needs to wait until it has fully constructed a type before putting it into the cache, even if it might be slower in some recursive cases. Better than wrong in them.~~
I don't know if clang is in the wrong here, so I'm rejecting a couple of statements. Its method does not allow for compilation in these circumstances, but may be technically permissible, even if gcc jives with how I envision templates operating.
Re:How to abuse a C++ compiler?
by pbsd on Thursday March 20, 2014 @12:41PM
(#7436717)
I don't see how that could not be a compiler error. The case b = 10 is more interesting because, as far as I can tell, it shows the differences between compilers that do two-phase lookup correctly and those that don't.
This seems to be a rare case where GCC is in the right, but I'd have to double-check the standard to be sure.
Re:How to abuse a C++ compiler?
by knome on Thursday March 20, 2014 @10:28PM
(#7440303)
It was pointed out to me after I joined #llvm when considering reporting this that using the value of "b", which is defined after "a" in the class before it has been initialized is, I suppose obviously in retrospect, undefined behavior.
In effect, I was subverting the undefined behavior associated with expanding a set of infinitely deep template instantiations by accessing a variable that was as yet undefined at the point of access.
So, my initial retractions stand, and this compiler behavior will remain appropriately undefined.
Re:How to abuse a C++ compiler?
by danbruc on Thursday March 20, 2014 @09:27AM
(#7435352)
The answer yes is not reasonable for Russell's paradox and therefore most likely not conforming, too. (Assuming the program really encodes Russell's paradox which I did not try to verify.)
UPDATE: Forget the part about conformance - if it is undefined behavior every result is of course correct. I made the wrong assumption that the result should definitely be an error if the correct behavior is not defined or can not be attained.
Re:How to abuse a C++ compiler?
by Tobu on Thursday March 20, 2014 @09:36AM
(#7435407)
Clang is mathematically wrong in all but the first case.
I'm not sure what it says about C++ conformance; the problem is likely Clang but I'm not sure how sound template metaprogramming or C++ templates are supposed to be.
Re:How to abuse a C++ compiler?
by Tobu on Thursday March 20, 2014 @07:05PM
(#7439498)
So the lack of soundness is in the C++ templates spec.
Nevertheless, I think it's bad form for a compiler to take full advantage of undefined behaviour. Reinterpreting undefined behaviour as “should not compile” (failing that, “must crash”) is possible for almost all of C99; CompCert (formally verified C compiler) obviously does it, and the implementation is the easy part (formal verification being the complicated part).
Re:How to abuse a C++ compiler?
by cbsmith on Thursday March 20, 2014 @10:10PM
(#7440255)
> So the lack of soundness is in the C++ templates spec.
That's not a lack of soundness. It's a lack of specificity. For cases like this, that's actually a feature, not a bug.
> Nevertheless, I think it's bad form for a compiler to take full advantage of undefined behaviour.
What exactly is "take full advantage of undefined behaviour"? Isn't any behaviour just as different from "undefined" as anything else?
> Reinterpreting undefined behaviour as “should not compile” (failing that, “must crash”) is possible for almost all of C99; CompCert (formally verified C compiler) obviously does it, and the implementation is the easy part (formal verification being the complicated part).
Clang not only has an ability to do this, it includes a flag to do a quick check specifically for undefined behaviour as well as an ability to provide a trap for undefined behaviour, which is terribly useful for tool builders.
Re:How to abuse a C++ compiler?
by aidenn0 on Friday March 28, 2014 @12:47AM
(#7485399)
So a compiler must emit a check for signed integer overflow on every integer operation? That seems absurd. It makes sense to optionally instrument for that (as clang offers) but is otherwise ridiculous.
Re:How to abuse a C++ compiler?
by unwind on Thursday March 20, 2014 @11:25AM
(#7436088)
C++'s template mechanism is not implemented by the preprocessor.
Templates control actual code generation (that's sort of their point, with generics) so they must be handled later in the compilation, by the compiler proper.
Re:How to abuse a C++ compiler?
by jasallen on Thursday March 20, 2014 @10:07AM
(#7435565)
as far as I can tell, the Russell's paradox example would be the same without the template "template<typename U, typename V>"?
in other words, both compilers seem to be being consistent with the previous demonstrated behavior using self-recursion, and the more generic template is being ignored altogether.
constexpr functions where quite restrictive in C++11. They could pretty much only contain a single expression and a return. C++14 significantly relaxes these restrictions.
http://en.wikipedia.org/wiki/C%2B%2B14#Relaxed_constexpr_res...
Type-level meta programming is still confined to using templates though...
https://github.com/hun-nemethpeter/cpp-reflector-mini/blob/m...
it calls a function at compile time to expand placeholders, and then compiles the generated string via the mixin() statement and carries on with the code. far more pleasant than c++.
Which there is less of need for this, since we have better type deduction in C++11 and 14 as well.
One of the fundamental properties people expect from a step called "compilation" is that it finishes. Also templates notation wasn't designed for this kind of use, and does not scale well for complex programs.
In Lisp code generation is done in Lisp, a notation that scales quite well, and there isn't a separarted step called "compilation". There is no general assumption that "execution" finishes.
It's more than a bit awkward really.
Still, the OP is right: it's funny.
For the MutualRec it may not set the static const value as being derived from a constexpr until after having retrieved the value, causing it to access the member in the cache before it's constexpr'ness has been determined.
For the Contra example, it would return 1 from having not'ed the 0 it found in the uninitialized static const.
Russel's paradox would be subverted in the same manner as the Contra example, returning true because it found false ( 0 ) in an uninitialized member variable after looking up the partially instatiated template instantiation in a cache.
This is all conjecture, but it fits, and it's where I would look first if debugging this. I do not know whether this is allowed by the standard, but I'd wager they did not account for referring cyclically to partially instantiated template parameterizations in a cache. GCC's errorful death is probably the correct action here, rather than clang's return of spurious values.
The same thing happens with Russell's Paradox, with the added twist that HasElement<M, M> always resolves to the specialization HasElement<M, T> (it 'fits' better) rather than the general version, which is never used.
~~clang needs to wait until it has fully constructed a type before putting it into the cache, even if it might be slower in some recursive cases. Better than wrong in them.~~
I don't know if clang is in the wrong here, so I'm rejecting a couple of statements. Its method does not allow for compilation in these circumstances, but may be technically permissible, even if gcc jives with how I envision templates operating.
This seems to be a rare case where GCC is in the right, but I'd have to double-check the standard to be sure.
In effect, I was subverting the undefined behavior associated with expanding a set of infinitely deep template instantiations by accessing a variable that was as yet undefined at the point of access.
So, my initial retractions stand, and this compiler behavior will remain appropriately undefined.
Edit: checked iso/iec 14882 (c++03), 14.7.1 Implicit instantiation. §14 "The result of an infinite recursion in instantiation is undefined".
UPDATE: Forget the part about conformance - if it is undefined behavior every result is of course correct. I made the wrong assumption that the result should definitely be an error if the correct behavior is not defined or can not be attained.
Nevertheless, I think it's bad form for a compiler to take full advantage of undefined behaviour. Reinterpreting undefined behaviour as “should not compile” (failing that, “must crash”) is possible for almost all of C99; CompCert (formally verified C compiler) obviously does it, and the implementation is the easy part (formal verification being the complicated part).
That's not a lack of soundness. It's a lack of specificity. For cases like this, that's actually a feature, not a bug.
> Nevertheless, I think it's bad form for a compiler to take full advantage of undefined behaviour.
What exactly is "take full advantage of undefined behaviour"? Isn't any behaviour just as different from "undefined" as anything else?
> Reinterpreting undefined behaviour as “should not compile” (failing that, “must crash”) is possible for almost all of C99; CompCert (formally verified C compiler) obviously does it, and the implementation is the easy part (formal verification being the complicated part).
Clang not only has an ability to do this, it includes a flag to do a quick check specifically for undefined behaviour as well as an ability to provide a trap for undefined behaviour, which is terribly useful for tool builders.
Templates control actual code generation (that's sort of their point, with generics) so they must be handled later in the compilation, by the compiler proper.
in other words, both compilers seem to be being consistent with the previous demonstrated behavior using self-recursion, and the more generic template is being ignored altogether.
Am I perhaps missing a reason it was included?