Remove stray comment at the end of tst_qexplicitlyshareddatapointer.cpp
Pick-to: 6.6
Change-Id: I31a6c38002e56e7c43e527864ba3d9324950079f
Reviewed-by: Santhosh Kumar <santhosh.kumar.selvaraj@qt.io>
QSpan is Qt's version of std::span. While we usually try not to
reimplement std functionality anymore, the situation is different with
QSpan. Spans are non-owning containers, so the usual impedance
mismatch between owning STL and Qt containers doesn't apply here:
QSpan implicitly converts to std::span and vice versa, making STL and
Qt APIs using spans completely interoperable.
We add QSpan mainly for two reasons: First, we don't want to wait
until we require C++20 in Qt and can use std::span. Second, in the
view of this author, some design decisions in std::span hurt the
primary use-case of spans: type-erasure for containers. This results
in two major deviations of QSpan from std::span: First, any rvalue
container is convertible to QSpan, allowing seamless passing of owning
containers to functions taking spans:
void sspan(std::span<T>);
void qspan(QSpan<T>);
std::vector<T> v();
sspan(v()); // ERROR: rvalue owning container
auto tmp = v();
sspan(tmp); // OK, lvalue
qspan(v()); // OK
This author believes that it's more helpful to have compilers and
static checkers warn about a particular wrong usage than to make
perfectly valid use-cases impossible or needlessly verbose to code.
The second deviation from std::span is that fixed-size span
constructors are also implicit. This isn't as clear-cut, because an
explicit QSpan{arg} isn't per-se bad. However, it means you can't
transparently change from a function taking decltype(arg) to one
taking QSpan and back. Since that's exactly what we intend to do in Qt
going forward, in the interest of source-compatibility, the ctors are
all implicit.
Otherwise, the API of QSpan follows the std::span API very
closely. Like std::span, QSpan isn't equality_comparable, because it's
not clear what equality means for spans (element-wise equal, or (ptr,
size)-wise equal?). The major API additions are Qt-ish versions of std
API functions: isEmpty() on top of empty() and sliced() instead of
subspan(). The (nullary) first()/last() functions (Qt speak for
front()/back()) clash with the std::span function templates of the
same name, so are not provided.
This patch adds QSpan as private API. We intend to make it public API
in the future.
Pick-to: 6.6
Fixes: QTBUG-108124
Change-Id: I3f660be90eb408b9e66ff9eacf5da4cba17212a6
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Dennis Oberst <dennis.oberst@qt.io>
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
We need deduction guides to turn the AtomicPointer template argument
(the pointee) into a pointer:
QAtomicPointer<int> → QAtomicScopedValueRollback<int*>
Extend a test to cover pointers, too.
Fixes: QTBUG-115105
Pick-to: 6.6 6.5
Change-Id: Ib416c6a43e4da480b707a0bf6a10d186bbaad163
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
It's a bit cumbersome, but works, in principle, using CTAD.
Pick-to: 6.6 6.5 6.2
Task-number: QTBUG-114200
Change-Id: Ib7354180e870a695a978edabf684aedfcf9d9ecc
Reviewed-by: Jarek Kobus <jaroslaw.kobus@qt.io>
Add the boilerplate standalone test prelude to each test, so that they
can be opened with an IDE without the qt-cmake-standalone-test script,
but directly with qt-cmake or cmake.
Boilerplate was added using the following scripts:
https://git.qt.io/alcroito/cmake_refactor
Manual adjustments were made where the code was inserted in the wrong
location.
Task-number: QTBUG-93020
Change-Id: I28b6d3815c5f43d2c33ea65764f6f3f8f129eaf3
Reviewed-by: Amir Masoud Abdol <amir.abdol@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
INTEGRITY has a pre-P1115 implementation of std::erase/erase_if that
returns void instead of the number of erased elements, so make q20's
implementation more specialized, so the compiler will pick it over
INTEGRITY's (Marc's idea from the code review).
Change-Id: I88d025a3f90cdd65f5bb73beb8a39e32ccf12d9b
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Restrict the permissible value_types to those QStringView can take,
plus QLatin1Char. All of these implicitly convert to QChar and give
the correct result, even when converted char-by-char.
Task-number: QTBUG-106198
Pick-to: 6.6
Change-Id: Icb44244cb08af391161c4309467d4e0d2d3d3d62
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Reviewed-by: Dennis Oberst <dennis.oberst@qt.io>
Implemented assign() methods for QByteArray to align with the
criteria of std::basic_string, addressing the previously missing
functionality. This is a subset of the overloads provided by the
standard.
Reference:
https://en.cppreference.com/w/cpp/string/basic_string/assign
[ChangeLog][QtCore][QByteArray] Added assign().
Fixes: QTBUG-106199
Change-Id: I899b14d74e8f774face8690303efb8610ead95b5
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Refactor the 'CHECK' macro to eliminate the capacity check and
explicitly verify that no reallocation occurred. The previous
implementation had to pass constants to suppress the issue arising
from differing growth rates between implementations.
Additionally, improve the 'std::stringstream' versions of the test
by incorporating the correct values. In the previous implementation,
the usage of:
auto tData = V(9);
~~~
std::stringstream ss("9 9 ");
had several issues. Firstly, it used the wrong test data since the
container's value_type of '(char) 9' resulted in a tab character '\t',
which was not accurately reflected in the stringstream assignment.
Secondly, this value caused problems in how stringstreams interprets it.
To address these issues, let's make the following improvements:
1. Use a default test value of 65 instead of (char) 9. This value, which
represents the character 'A', is less likely to cause errors and is more
intuitive.
2. Use the tData variable for the assignments in the stringstream. This
ensures that the correct data from the container is used.
3. Change the test value between the assign() calls to verify that the
container's contents are successfully overwritten.
These changes ensure, that the test cases are more accurate and
reliable.
Amends: 3b0536bbe8.
Change-Id: I9441c4818106bf93e93a1a5d2d2d54c89d80e7b0
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
While std::vector::assign() returns void, std::basic_string::assign()
returns std::basic_string&. In Qt, we want to be consistent between
{QVLA,QList,QString,QByteArray}::assign(), and returning *this is the
more general solution, so do that.
Task-number: QTBUG-106196
Task-number: QTBUG-106200
Change-Id: I2689b4af032ab6fb3f8fbcb4d825d5201ea5abeb
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Implemented assign() methods for QList to align with the criteria of
std::vector, addressing the previously missing functionality.
Reference:
https://en.cppreference.com/w/cpp/container/vector/assign
[ChangeLog][QtCore][QList] Added assign().
Fixes: QTBUG-106196
Change-Id: I5df8689c020dafde68d2cd7d09c769744fa8f137
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
This file, like the majority of qtbase, uses a space between template
and the opening of the template argument list. Add it.
Change-Id: I927cb2b1b9620ae108e913343d995373493e8981
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
This is handled by the Objective-C runtime nowadays, where it will
abort if the situation is detected, with the option to break on
objc_autoreleasePoolInvalid to debug the situation.
Pick-to: 6.5
Change-Id: Idf2c4aacc77e41a3deebf270303f4f13cfb0819b
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
The options included by --help-all, although they are "specific to
Qt", are "specific" to all Qt applications, so - in the present
context, of QCommandLineParser - not specific at all. It's the options
described by -h that are specific, to the present command; the Qt
options are generic (in the present context).
So rework the help string for --help-all itself and the documentation
of the function. It had, in any case, an overly-complex first line,
that descended into too much detail. Updated test to match.
Pick-to: 6.5
Task-number: QTBUG-111228
Change-Id: I06da0af41be60e6e1b7616984001ddb9ca33aad6
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: David Faure <david.faure@kdab.com>
The parameter passed to reserve() is just a hint. The container
implementation is free to choose a larger capacity, and some do
(e.g. QList in prepend optimization mode).
Fix the test by querying the container for its post-make<>()
capacity() and taking a larger-than-expected initial capacity() into
account when later re-checking the capacity().
Change-Id: Id8f26f14e8df9d685ca2387ec4a52d74fea7cb9d
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
It took me a sec to figure out the relation between the comment and
the code line following it. Make it easier for the next guy and add a
bit more infos.
Amends 7cbdc8abbd.
Change-Id: I4ff2d9a52aef643a92339df32cc86f686a689a9a
Reviewed-by: Dennis Oberst <dennis.oberst@qt.io>
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
We use a single line per test slot everywhere else, ignoring even
line-length limitations, to keep the function names aligned for easier
parsing.
Amends 7cbdc8abbd.
Change-Id: Iaf2941aae88392d407d688fc4a7537fcdc0a5851
Reviewed-by: Dennis Oberst <dennis.oberst@qt.io>
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
STL algorithms, in general, don't specify how often the function
objects passed to them are copied during the run of the
algorithm.
While generate_n is above any reasonable suspicion of copying the
function object after the first invocation, passing a mutable lambda
containing the counter is still an anti-pattern we don't want people
to copy.
Fix in the usual way, by keeping the counter external to the lambda.
As a drive-by, replace post- with pre-increment.
Amends dc091e7443.
Pick-to: 6.5 6.2
Change-Id: I9c44e769fd41e5f7157179a2be4c3534424cf913
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Removing a few unused variables in auto tests that were triggering
`-Wunused-but-set-variable`.
Pick-to: 6.5
Change-Id: I74bd0d7335d8bddeb18687b18c8a8be965f9fa20
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
QMultiHash has access to two sizes: one of them is shared with QHash,
stored in QHashPrivate::Data::size, which counts keys; the other, which
is what our public size() function returns, is stored in
QMultiHash::m_size and counts plain (key,value) entries. We forgot to
update it in the non-const operator[] that created a node.
I've reviewed the rest of the code and can't find any more places where
the item count may be changed and m_size isn't updated.
[ChangeLog][QtCore][QMultiHash] Fixed a bug that caused an element that
was created by operator[] to not be counted, resulting in a hash map
with an incorrect element count and which could cause an assertion
failure depending on how the hash was later mutated.
Fixes: QTBUG-112534
Pick-to: 6.2 6.4 6.5
Change-Id: Idd5e1bb52be047d7b4fffffd17527ec274e1d99e
Reviewed-by: Lars Knoll <lars@knoll.priv.no>
In C++17, unqualified lookup doesn't find function templates that
require ADL from a call with explicit template arguments, unless
another function template of that name is in scope (otherwise, the <
is parsed as operator less-than instead).
P0846, merged for C++20, fixes this to repeat the name lookup, parsing
the < as indicating a template.
We have API in Qt (Tuple Protocol for some types, e.g. QPoint) that
work for the purpose of Structured Bindings, but don't work for manual
unqualified calls when P0846 semantics are missing, and we're adding
more, to QVariant, so add a macro to handle the issue.
The macro simply declares a function template overload of the given
name for a throw-away struct, thereby bringing, for that one name,
P0846 semantics into C++17.
When we require C++20, we can drop this again.
Amends:
- fb6b7869e8 for QPoint(F)
- 8ae9431c79 for QMargins(F)
- 0e22001a3b for the rest
[ChangeLog][QtCore][QSize/F, QMargins/F, QPoint/F] Fixed manual
get<I>() calls (Tuple Protocol) in C++17 mode.
Task-number: QTBUG-111598
Change-Id: I2ffaef12c5bb6d82f75ce78a7c03c6789dfa0691
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
We test setKey() in repeated_setKey() these days, so speed up the test
of the test suite ever so slightly by passing the key to the ctor
instead of an explicit setKey() call.
Pick-to: 6.5
Change-Id: Ia2378c0f59cbfa9d95a0f3665b06655332247e2c
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
When this code was using QByteArray, whose data() is never nullptr,
the strings presented to the underlying C APIs were always valid NTSs
(nullptr is not a valid NTS).
With QByteArrayView, or with QT5_NULL_STRINGS != 1, this is no longer
the case. Check that all implementations are fine with that.
Pick-to: 6.5 6.4
Change-Id: I78251288a4784440af4a2daf095aed7c53867287
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
QCryptographicHash is move-only these days, so
QMessageAuthenticationCode should not be left behind.
[ChangeLog][QtCore][QMessageAuthenticationCode] Added move
constructor, move assignment operator and swap() function.
Fixes: QTBUG-111677
Change-Id: I420f24c04828e8ad7043a9e8c9e7e2d47dd183e0
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Previous setup of the test was failing in minimal static build if built
using the unity build because of the explicit inclusion of the qtcore
source files. In order to resolve this, I removed the inclusion of
qtcore's headers and made the test private.
Pick-to: 6.5
Task-number: QTBUG-109394
Change-Id: Id1c7b3b65ca2078354c235a718ff3e93a65362e6
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
All the other overloads are implemented using the new one.
Windows change relies on the pre-check in the code review making sure it
compiles.
[ChangeLog][QtCore][QThread] Added sleep(std::chrono::nanoseconds)
overload.
Task-number: QTBUG-110059
Change-Id: I9a4f4bf09041788ec9275093b6b8d0386521e286
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
state->rate is always larger than or equal to state->bitsInQueue;
when bitsInQueue == rate the queue is consumed and bitsInQueue is set to
0 again.
Done-with: Marc Mutz <marc.mutz@qt.io>
Pick-to: 6.5.0 6.5 6.4.3 6.4 6.2 5.15
Change-Id: I56d268a19fb3cd542cc027edc962253f09d97a14
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
... to make large data usable from other test functions.
Pick-to: 6.5 6.5.0 6.4 6.4.3 6.2
Change-Id: I302070121a8bb49f373c7711bc3ab9e6418874ef
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Also add one for types that are neither copy- nor move-constructible.
In contrast to resize(n), the QVLA(n) ctor worked for such types, so
make sure it stays that way.
Pick-to: 6.5 6.4 6.4.3 6.2 5.15
Change-Id: If54fbc9dd6a4808175c4bcb0ffb492b33c879746
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Use it in a few places.
[ChangeLog][QtCore][QMessageAuthenticationCode] Added
QCryptographicHash-style resultView().
Change-Id: I745d71f86f9c19c9a9aabb2021c6617775dab1cf
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Add Algorithm::NumAlgorithms and use it to iterate over all
statically-available algorithms, querying their hashLengthInternal().
This avoids having to statically_assert(<= MaxHashLength) everywhere,
and auto-adjusts the buffer size in SHA1_ONLY builds.
Yes, the extra case labels for NumAlgorithms are a nuisance, but at
least the compiler will remind us when we forget, unlike a missing
static_cast(<= MaxHashLength) that might easily be forgotten.
Adjust the test (which iterates over the QMetaEnum for
QCryptographicHash::Algorithm, so finds NumAlgorithms and tries to
pass it to the hash() function which responds with a
Q_UNREACHABLE(). Only test hashLength() == 0 for that enum value.
Pick-to: 6.5
Change-Id: I70155d2460464f0b2094e136eb6bea185effc9d5
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It was confusing entry capacity with the bucket capacity. The value
maxNumBuckets() returned was the maximum number of entries. This issue
was harmless: we would just fail to cap the maximum to an allocatable
size. But the array new[] in the Data constructors would have capped the
maximum anyway (by way of throwing std::bad_alloc).
So instead of trying to calculate what the maximum bucket count is so we
can cap at that, simplify the calculation of the next power of 2 while
preventing it from overflowing in our calculations. We continue to rely
on new[] throwing when we return count that is larger than the maximum
allocatable.
This commit changes the load factor for QHashes containing exactly a
number of elements that is exactly a power of two. Previously, it would
be loaded at 50%, now it's at 25%. For this reason, tst_QSet::squeeze
needed to be fixed to depend less on the implementation details.
Pick-to: 6.5
Change-Id: I9671dee8ceb64aa9b9cafffd17415f3856c358a0
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
I don't begin to understand the semantics of the trackers here, but
whatever they are, they break with the fallback std::swap() 3-moves
implementation and lose track of alive objects, so provide an ADL swap
that does the right thing.
Amends dd58ddd5d9 (I think).
Pick-to: 6.5 6.4 6.2 5.15
Change-Id: I1cd49c95dca2d103a26c2c7ac0a896929135a6c8
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Implemented assign() methods for QVarLengthArray to align
with the criteria of std::vector, addressing the previously
missing functionality.
Reference:
https://en.cppreference.com/w/cpp/container/vector/assign
[ChangeLog][QtCore][QVarLengthArray] Added assign().
Fixes: QTBUG-106200
Change-Id: If671069808ff561b0f4c77b6c7f7aca360a0c663
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
The growBy() function takes the _increment_ of the size(), so needs to
add size() to increment for the call to realloc().
Add a test which hangs (vanilla build) or explodes (valgrind build)
without the fix.
Amends 26b227e128.
Done-with: Eirik Aavitsland <eirik.aavitsland@qt.io>
Pick-to: 6.5 6.4
Fixes: QTBUG-110412
Change-Id: I7ea91342fdcb779825c88013a3f86ba6d90ef530
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Despite being move-only, std::vector<unique_ptr> advertizes
is_copyable:
https://quuxplusone.github.io/blog/2020/02/05/vector-is-copyable-except-when-its-not/
Our combined reallocation and resizing function, reallocate_impl(),
runs afoul of this when it uses std::is_copyable in a constexpr-if to
implement resize(n, v) without running into problems with move-only
types: the trait is true, but actual instantation runs into a
static_assert in the STL implementation.
To fix, move the problematic resize functionality out of
reallocate_impl() and into the resp. resize_impl overloads. The shrink
functionality remains in reallocate_impl(), because there are many
more users, and it only requires destructible<T>, which isn't
constraining at all.
Amends a00a1d8806.
Fixes: QTBUG-109745
Pick-to: 6.5 6.4
Change-Id: Ibc5b9cf5375108eb3d8f6c8a16d4fd02dadd73b1
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Just describe the row instead. We'd lose the original input in case of
failure, so I added a class to print that value on destruction. Example:
FAIL! : tst_QAlgorithms::countLeading64(0) Compared values are not the
same
Actual (qCountLeadingZeroBits(value)): 63
Expected (expected) : 64
Loc: [tests/auto/corelib/tools/qalgorithms/tst_qalgorithms.cpp(374)]
QWARN : tst_QAlgorithms::countLeading64(0) Original value was 0x1
Pick-to: 6.4 6.5
Fixes: QTBUG-109958
Change-Id: I69ecc04064514f939896fffd1738b1119cd80cf8
Reviewed-by: Dimitrios Apostolou <jimis@qt.io>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
In preparation to their deprecation / removal.
Change-Id: Ia073a9f7caabbc06063a1e416b23cdb12788b283
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
For QSet, the key_type is const, so can't test assiging to it.
Pick-to: 6.4
Change-Id: I9d363ef3fe52646b937d6a422227b19c48fdaf1f
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
This is a semantic patch using ClangTidyTransformator as in
qtbase/df9d882d41b741fef7c5beeddb0abe9d904443d8, but extended to
handle typedefs and accesses through pointers, too:
const std::string o = "object";
auto hasTypeIgnoringPointer = [](auto type) { return anyOf(hasType(type), hasType(pointsTo(type))); };
auto derivedFromAnyOfClasses = [&](ArrayRef<StringRef> classes) {
auto exprOfDeclaredType = [&](auto decl) {
return expr(hasTypeIgnoringPointer(hasUnqualifiedDesugaredType(recordType(hasDeclaration(decl))))).bind(o);
};
return exprOfDeclaredType(cxxRecordDecl(isSameOrDerivedFrom(hasAnyName(classes))));
};
auto renameMethod = [&] (ArrayRef<StringRef> classes,
StringRef from, StringRef to) {
return makeRule(cxxMemberCallExpr(on(derivedFromAnyOfClasses(classes)),
callee(cxxMethodDecl(hasName(from), parameterCountIs(0)))),
changeTo(cat(access(o, cat(to)), "()")),
cat("use '", to, "' instead of '", from, "'"));
};
renameMethod(<classes>, "count", "size");
renameMethod(<classes>, "length", "size");
except that the on() matcher has been replaced by one that doesn't
ignoreParens().
a.k.a qt-port-to-std-compatible-api V5 with config Scope: 'Container'.
Added two NOLINTNEXTLINEs in tst_qbitarray and tst_qcontiguouscache,
to avoid porting calls that explicitly test count().
Change-Id: Icfb8808c2ff4a30187e9935a51cad26987451c22
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
qhashfunctions.h defines a catch-all 2-arguments qHash(T, seed)
in order to support datatypes that implement a 1-argument overload
of qHash (i.e. qHash(Type)). The catch-all calls the 1-argument
overload and XORs the result with the seed.
The catch-all is constrained on the existence of such a 1-argument
overload. This is done in order to make the catch-all SFINAE-friendly;
otherwise merely instantiating the catch-all would trigger a hard error.
Such an error would make it impossible to build a type trait that
detects if one can call qHash(T, size_t) for a given type T.
The constraint itself is called HasQHashSingleArgOverload and lives in a
private namespace.
It has been observed that HasQHashSingleArgOverload misbehaves for
some datatypes. For instance, HasQHashSingleArgOverload<int> is actually
false, despite qHash(123) being perfectly callable. (The second argument
of qHash(int, size_t) is defaulted, so the call *is* possible.)
--
Why is HasQHashSingleArgOverload<int> false?
This has to do with how HasQHashSingleArgOverload<T> is implemented: as
a detection trait that checks if qHash(declval<T>()) is callable.
The detection itself is not a problem. Consider this code:
template <typename T>
constexpr bool HasQHashSingleArgOverload = /* magic */;
class MyClass {};
size_t qHash(MyClass);
static_assert(HasQHashSingleArgOverload<MyClass>); // OK
Here, the static_assert passes, even if qHash(MyClass) (and MyClass
itself) were not defined at all when HasQHashSingleArgOverload was
defined.
This is nothing but 2-phase lookup at work ([temp.dep.res]): the
detection inside HasQHashSingleArgOverload takes into account the qHash
overloads available when HasQHashSingleArgOverload was declared, as well
as any other overload declared before the "point of instantiation". This
means that qHash(MyClass) will be visible and detected.
Let's try something slightly different:
template <typename T>
constexpr bool HasQHashSingleArgOverload = /* magic */;
size_t qHash(int);
static_assert(HasQHashSingleArgOverload<int>); // ERROR
This one *does not work*. How is it possible? The answer is that 2-phase
name lookup combines the names found at definition time with the names
_found at instantiation time using argument-dependent lookup only_.
`int` is a fundamental type and does not participate in ADL. In the
example, HasQHashSingleArgOverload has actually no qHash overloads to
even consider, and therefore its detection fails.
You can restore detection by moving the declaration of the qHash(int)
overload *before* the definition of HasQHashSingleArgOverload, so it's
captured at definition time:
size_t qHash(int);
template <typename T>
constexpr bool HasQHashSingleArgOverload = /* magic */;
static_assert(HasQHashSingleArgOverload<int>); // OK!
This is why HasQHashSingleArgOverload<int> is currently returning
`false`: because HasQHashSingleArgOverload is defined *before* all the
qHash(fundamental_type) overloads in qhashfunctions.h.
--
Now consider this variation of the above, where we keep the qHash(int)
overload after the detector (so, it's not found), but also prepend an
Evil class implicitly convertible from int:
struct Evil { Evil(int); };
size_t qHash(Evil);
template <typename T> constexpr bool HasQHashSingleArgOverload = /* magic */;
size_t qHash(int);
static_assert(HasQHashSingleArgOverload<int>); // OK
Now the static_assert passes. HasQHashSingleArgOverload is still not
considering qHash(int) (it's declared after), but it's considering
qHash(Evil). Can you call *that* one with an int? Yes, after a
conversion to Evil.
This is extremely fragile and likely an ODR violation (if not ODR, then
likely falls into [temp.dep.candidate/1]).
--
Does this "really matter" for a type like `int`? The answer is no. If
HasQHashSingleArgOverload<int> is true, then a call like
qHash(42, 123uz);
will have two overloads in its overloads set:
1) qHash(int, size_t)
2) qHash(T, size_t), i.e. the catch-all template. To be pedantic,
qHash<int>(const int &, size_t), that is, the instantiation of the
catch-all after template type deduction for T (= int)
([over.match.funcs.general/8]).
Although it may look like this is ambiguous as both calls have perfect
matches for the arguments, 1) is actually a better match than 2) because
it is not a template specialization ([over.match.best/2.4]).
In other words: qHash(int, size_t) is *always* called when the argument
is `int`, no matter the value of HasQHashSingleArgOverload<int>. The
catch-all template may be added or not to the overload set, but it's
a worse match anyways.
--
Now, let's consider this code:
enum MyEnum { E1, E2, E3 };
qHash(E1, 42uz);
This code compiles, although we do not define any qHash overload
specifically for enumeration types (nor one is defined by MyEnum's
author).
Which qHash overload gets called? Again there are two possible
overloads available:
1) qHash(int, size_t). E1 can be converted to `int` ([conv.prom/3]),
and this overload selected.
2) qHash(T, size_t), which after instantiation, is qHash<MyEnum>(const
MyEnum &, size_t).
In this case, 2) is a better match than 1), because it does not require
any conversion for the arguments.
Is 2) a viable overload? Unfortunately the answer here is "it depends",
because it's subject to what we've learned before: since the catch-all
is constrained by the HasQHashSingleArgOverload trait, names introduced
before the trait may exclude or include the overload.
This code:
#include <qhashfunctions.h>
enum MyEnum { E1, E2, E3 };
qHash(E1, 42uz);
static_assert(HasQHashSingleArgOverload<MyEnum>); // ERROR
will fail the static_assert. This means that only qHash(int, size_t) is
in the overload set.
However, this code:
struct Evil { Evil(int); };
size_t qHash(Evil);
#include <qhashfunctions.h>
enum MyEnum { E1, E2, E3 };
qHash(E1, 42uz);
static_assert(HasQHashSingleArgOverload<MyEnum>); // OK
will pass the static_assert. qHash(Evil) can be called with an object of
type MyEnum after an user-defined conversion sequence
([over.best.ics.general], [over.ics.user]: a standard conversion
sequence, made of a lvalue-to-rvalue conversion + a integral promotion,
followed by a conversion by constructor [class.conv.ctor]).
Therefore, HasQHashSingleArgOverload<MyEnum> is true here; the catch-all
template is added to the overload set; and it's a best match for the
qHash(E1, 42uz) call.
--
Is this a problem? **Yes**, and a huge one: the catch-all template does
not yield the same value as the qHash(int, size_t) overload. This means
that calculating hash values (e.g. QHash, QSet) will have different
results depending on include ordering!
A translation unit TU1 may have
#include <QSet>
#include <Evil>
QSet<MyEnum> calculateSet { /* ... */ }
And another translation unit TU2 may have
#include <Evil>
#include <QSet> // different order
void use() {
QSet<MyEnum> set = calculateSet();
}
And now the two TUs cannot exchange QHash/QSet objects as they would
hash the contents differently.
--
`Evil` actually exists in Qt. The bug report specifies QKeySequence,
which has an implicit constructor from int, but one can concoct infinite
other examples.
--
Congratulations if you've read so far.
=========================
=== PROPOSED SOLUTION ===
=========================
1) Move the HasQHashSingleArgOverload detection after declaring the
overloads for all the fundamental types (which we already do anyways).
This means that HasQHashSingleArgOverload<fundamental_type> will now
be true. It also means that the catch-all becomes available for all
fundamental types, but as discussed before, for all of them we have
better matches anyways.
2) For unscoped enumeration types, this means however an ABI break: the
catch-all template becomes always the best match. Code compiled before
this change would call qHash(int, size_t), and code compiled after this
change would call the catch-all qHash<Enum>(Enum, size_t); as discussed
before, the two don't yield the same results, so mixing old code and new
code will break.
In order to restore the old behavior, add a qHash overload for
enumeration types that forwards the implementation to the integer
overloads (using qToUnderlying¹).
(Here I'm considering the "old", correct behavior the one that one gets
by simply including QHash/QSet, declaring an enumeration and calling
qHash on it. In other words, without having Evil around before including
QHash.)
This avoids an ABI break for most enumeration types, for which one
does not explicitly define a qHash overload. It however *introduces*
an ABI break for enumeration types for which there is a single-argument
qHash(E) overload. This is because
- before this change, the catch-all template was called, and that
in turn called qHash(E) and XOR'ed the result with the seed;
- after this change, the newly introduced qHash overload for
enumerations gets called. It's very likely that it would not give
the same result as before.
I don't have a solution for this, so we'll have to accept the ABI
break.
Note that if one defines a two-arguments overload for an enum type,
then nothing changes there (the overload is still the best match).
3) Make plans to kill the catch-all template, for Qt 7.0 at the latest.
We've asked users to provide a two-args qHash overload for a very long
time, it's time to stop working around that.
4) Make plans to switch from overloading qHash to specializing std::hash
(or equivalent). Specializations don't overload, and we'd get rid of
all these troubles with implicit conversions.
--
¹ To nitpick, qToUnderlying may select a *different* overload than
the one selected by an implicit conversion.
That's because an unscoped enumeration without a fixed underlying type
is allowed to have an underlying type U, and implicitly convert to V,
with U and V being two different types (!).
U is "an integral type that can represent all the enumerator values"
([dcl.enum/7]). V is selected in a specific list in a specific order
([conv.prom]/3). This means that in theory a compiler can take enum E {
E1, E2 }, give it `unsigned long long` as underlying type, and still
allow for a conversion to `int`.
As far as I know, no compiler we use does something as crazy as that,
but if it's a concern, it needs to be fixed.
[ChangeLog][Deprecation Notice] Support for overloads of qHash with only
one argument is going to be removed in Qt 7. Users are encouraged to
upgrade to the two-arguments overload. Please refer to the QHash
documentation for more information.
[ChangeLog][Potentially Binary-Incompatible Changes] If an enumeration
type for which a single-argument qHash overload has been declared is
being used as a key type in QHash, QMultiHash or QSet, then objects of
these types are no longer binary compatible with code compiled against
an earlier version of Qt. It is very unlikely that such qHash overloads
exist, because enumeration types work out of the box as keys Qt
unordered associative containers; users do not need to define qHash
overloads for their custom enumerations. Note that there is no binary
incompatibity if a *two* arguments qHash overload has been declared
instead.
Fixes: QTBUG-108032
Fixes: QTBUG-107033
Pick-to: 6.2 6.4
Change-Id: I2ebffb2820c553e5fdc3a341019433793a58e3ab
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The "qhash" test relied on the fact that those four elements would
produce a different order with a zero and a non-zero seed. But since
commit b057e32dc4 removed the setting of a
deterministic non-zero seed, this test had a 1 in 4! chance of failing.
Since 4! = 24, 128 retries should be more than enough to ensure we do
find at least hash seed that provokes a different order.
Fixes: QTBUG-107725
Change-Id: I3c79b7e08fa346988dfefffd171ee61b79ca5489
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
