Before this change, the report presented to the user would be
completely silent about them. For the sake of consitency, register each
file before trying to compile it. This will then add an entry for empty
files and modules.
Task-number: QTBUG-124667
Pick-to: 6.8
Change-Id: I502660b7a16a67a173763f9ea2b081cbcceb5658
Reviewed-by: Sami Shalayel <sami.shalayel@qt.io>
It is possible to pass --only-bytecode to qmlcachegen. As the name
implies, this only generates the bytecode for the qml files and does not
compile them. This case was not taken into account which could lead to
files commands depend upon not being generated.
Therefore, keep track of empty and only-bytecode modules in files
generated by cmake and pass them to qmlaotstats upon aggregation such
that it can include that information in the report.
Also, only pass the arguments specific to aotstats to qmlcachegen if
--only-bytecode is not set for that module.
Fixes: QTBUG-130084
Task-number: QTBUG-124667
Pick-to: 6.8
Change-Id: I44b4a80e8a6fd2f9bc16ae1bb2c8d540ff3b697b
Reviewed-by: Sami Shalayel <sami.shalayel@qt.io>
Previously, it would hold the location of the copy in the build
directory or in the resource file system.
This should improve the accuracy of the information that we can report
back. For example, the file location of files reported in aotstats will
now be those in the source folder.
The changes necessary for this patch were found through trial and error
by asserting that the filePath respect a certain form and fixing what I
could. There remain cases where the filePath still contains a location
in the build directory. We would need some other change to report those
accurately as well.
Finally, sometimes the filePath is empty or does not contain a valid
path at all but instead what seems to be the location of an included
file. It can have the form "private/..._p.h" for example.
Pick-to: 6.8
Change-Id: I77a2596f6c4a179580d924a6e9324e43dbb86c0c
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We set those options from CMake. qmlcachegen can still dump (empty) AOT
stats if it's compiling only to byte code, and it's not wrong to still
tell it the module ID.
Amends commit 6ae36202fd
Change-Id: I3990c0783f9fdf0c7c7263675dc82c8c74851eb8
Reviewed-by: Olivier De Cannière <olivier.decanniere@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
[ChangeLog][QtQml] qmlcachegen will now exit with a failure state if
--only-bytecode is set as the same time as Script Compiler exclusive
options. The only-bytecode flag skips the compilation altogether.
Remove one of the conflicting flag to solve the issue.
Change-Id: I5c6a0610ec4c2eb06624c2d123ab62bc633a46b8
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
This patch introduces the collection of statistics about the
ahead-of-time compilation of functions and bindings to Cpp by
qmlcachegen. This is done by having qmlcachegen save an aotstats file
for every qml file it compiles. This file contains, for every function
and binding, whether the Cpp codegen was successful, its duration and a
potential error message
Task-number: QTBUG-124667
Change-Id: Iba9a72be04f6642688533a3ae12ea687296c85e1
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We want qmlcachegen to accept the same arguments as qmlsc so that people
don't have to fiddle with the build system when they need to switch
between them.
Pick-to: 6.7 6.5
Change-Id: If21cbd262104c68c96649b4b0882c079c3862f05
Reviewed-by: Sami Shalayel <sami.shalayel@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
[ChangeLog][qmlcachegen] Added the --warnings-are-errors option that
treats warnings as errors when generating C++ code.
Like gcc's -Werror, this option is useful when aiming at warning-free
QML code. Add it to the QT_QMLCACHEGEN_ARGUMENTS target property to
error out once warnings have been encountered.
Change-Id: If72875b33052b37642b18fa192a1344c347990d8
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
The function prolog logic is now separated in its own basic block. The
first "real" block with user code starts at offset 0.
Having the function prolog as a hidden part of the first block caused
some inconsistencies in block generation and would create empty blocks.
This happened for example when a back edge of a loop would target offset
0 in code where a loop condition is the very first set of instructions
that are run. This is because the target block offset didn't exist due
to it being part of the hidden prolog block.
Validation for the basic blocks was also added. This checks for three
things at the moment:
1. That return and throw blocks don't have jump targets.
2. That the basic blocks graph is connected.
3. That jump targets are the first offset of a block.
Test tst_QmlCppCodegen::basicBlocksWithBackJump_infinite() is expected
to fail because it contains an infinite loop and the basic blocks that
are generated for it are inconsistent due to dead-code elimination
happening earlier in compilation.
Debug outputs for dumping basic blocks were also adapted to reflect
these changes.
Change-Id: I513f73856412d488d443c2b47a052b0023d45496
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
This replaces the logging category magic to determine whether we should
output warnings.
Change-Id: Ibc95ed70227e3438504e921693aa4af9a1d0376e
Reviewed-by: Sami Shalayel <sami.shalayel@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
There are many ways to "hide" the qmldir from the engine at run time,
which turns singletons into regular types. While all of this is invalid,
we should not assert on it, but rather produce a legible warning.
Furthermore, sharpen the importing of extra modules from qrc as implicit
imports. We should really only import modules the file in question can
ever be part of. Otherwise we needlessly produce the above situation and
hide legitimate warning messages.
Amends commit 7517c1b3ae.
Now we need to teach our tools about the default import paths in the
resorurce file system. They cannot guess any type they may find in any
resource file anymore.
Pick-to: 6.5
Task-number: QTBUG-106929
Change-Id: Ic8c02396d10830a7f461e8a81649bb8c9a1add1f
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Sami Shalayel <sami.shalayel@qt.io>
We want to trigger the qFatal() when running qmlcachegen or qmlsc, but
we don't want to trigger it in qmllint, even if we are using the same
QtMsgType for the severity of the messages.
In turn, demote qmlCompiler messages to QtWarningMsg in qmlcachegen.
Pick-to: 6.4
Task-number: QTBUG-107168
Change-Id: Ib660df41742b2d426241eb29ac4c91f4933c5ba1
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
This is a semantic patch using ClangTidyTransformator as in
qtbase/df9d882d41b741fef7c5beeddb0abe9d904443d8:
auto QtContainerClass = anyOf(
expr(hasType(cxxRecordDecl(isSameOrDerivedFrom(hasAnyName(classes))))).bind(o),
expr(hasType(namedDecl(hasAnyName(<classes>)))).bind(o));
makeRule(cxxMemberCallExpr(on(QtContainerClass),
callee(cxxMethodDecl(hasAnyName({"count", "length"),
parameterCountIs(0))))),
changeTo(cat(access(o, cat("size"), "()"))),
cat("use 'size()' instead of 'count()/length()'"))
a.k.a qt-port-to-std-compatible-api with config Scope: 'Container',
with the extended set of container classes recognized.
Change-Id: Idb1f75dfe2323bd1d9e8b4d58d54f1b4b80c7ed7
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We need those to prevent usage of the host Qt's import path when
building QML files for the target.
Task-number: QTBUG-106365
Pick-to: 6.4 6.4.0
Change-Id: I34499af850fb759603206cb429ff97633c76054d
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
This change makes qmljslogger use an ID based system for categorizing
logging entries instead of using an enum. This allows plugins to
register their own logging categories after the fact.
It's also necessary for us to later show the warning ID when
printing warnings and for creating documentation for each ID entry.
Currently not every ID maps cleanly to only one type of warning,
this has to be cleaned up in a follow-up change.
Task-number: QTBUG-103453
Change-Id: I4cac6be7ca165b938e0ea032d077823bf17baf75
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Replace the current license disclaimer in files by
a SPDX-License-Identifier.
Files that have to be modified by hand are modified.
License files are organized under LICENSES directory.
Pick-to: 6.4
Task-number: QTBUG-67283
Change-Id: I63563bbeb6f60f89d2c99660400dca7fab78a294
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
Converts qmlcompiler into a shared library so it isn't duplicated when
statically linked in tools and so that plugins can link against it.
Change-Id: I91e13cc588796f71a5463dbdce21e42a74120565
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
We've accidentally disabled the pragma Strict check when re-designing
the logging system.
Change-Id: I74cf0d4a45742831037ec0bc0615f1f6a15b7023
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
High-level goal: be able to reuse existing infrastructure
"as is" to configure semantic analysis categories in tools
(qmllint, qmltc, qmlsc, etc.)
To achieve that, simplify the logging to always "log"
something, without explicitly specifying the severity. The
severity is now baked into the category (and we can extend
those to cover different cases)
One slight deviation is the cache generation which likes
to do its own thing at present. Provide a "forced logging"
option where we can specify which severify we want. The
hope is that this gets removed at some point
Particular list of (noteworthy) changes:
* No more "thresholding" by the level (this is rarely needed
and is actually questionable). Instead, we can ignore a
particular category explicitly
* Category levels are repurposed as category severities
(at least from the high-level picture that always should've
been this way)
* log{Warning,Info,Critical} removed. We use category severity
instead
* "category error" makes zero sense so removed: if our severity
is:
- QtWarningMsg (qmllint), it is already an "error"
- QtCriticalMsg (compilers), it is already an "error"
* Align m_output and m_{infos,warnings,errors} stored information
* Accept the fact that we don't support QtDebugMsg and QtFatalMsg
* Additional categories added to cover for places where the same
category would be both an error and not an error
Task-number: QTBUG-100052
Change-Id: I3cd5d17d58be204f48428877bed053f756ac40a8
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
This change adds a warning if we encounter a component marked as a
singleton in a qmldir file which does not contain a pragma
Singleton entry and vice versa.
Note that the warning only gets triggered if the singleton is
actually used.
Fixes: QTBUG-98558
Change-Id: Id7c63f48ba49759c15dffcaee0270c7caab2eb7d
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
The tools will still grudgingly accept qmltypes files being passed via
the -i option. We generally expect qmldir files, though. Ignoring the
qmldir file and importing the qmltypes directly, ignores qmldir imports,
dependencies and other component entries. This leads to unresolvable
types.
[ChangeLog][QML Tooling] qmllint expects qmldir files, not qmltypes
files to be passed via the -i option now. This enables it to see the
imports and dependencies of the module being imported. For backwards
compatibility it still accepts qmltypes files, with a warning.
Pick-to: 6.3
Fixes: QTBUG-99043
Change-Id: I5ed32d7e78df1e604aaf1bfa2ebda09d5d57b628
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
qmlcachegen compiles bindings and functions to C++ as far as
QQmlJSAotCompiler can. It does respect "pragma Strict" and rejects the
file if it's violated. Furthermore, it sets up the logger to follow the
qt.qml.compiler.aot logging category. By default it's completely silent.
Compiling the examples with qmlcachegen exposes a bug in the type
resolver where it returns an invalid generic type. It should never do
that. Fix it by returning JSValue.
[ChangeLog][QtQml][Important Behavior Changes] QML bindings and
functions are now compiled to C++ by qmlcachegen, if possible. Use the
qt.qml.compiler.aot logging category to receive diagnostics about the
compilation.
Task-number: QTBUG-98305
Change-Id: I6953812c3fd20b68339617a5714fcbe16a384360
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We don't need --qmljs-runtime after all. However, we do want to pass
custom arguments to qmlcachegen. This way we don't need to change the
world whenever we invent one.
Pick-to: 6.2
Change-Id: I99f0c99c4b15e469605dc43d26871fa43dc36c08
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
Reviewed-by: Fawzi Mohamed <fawzi.mohamed@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
'resource-file-mapping option' is actually using '=' in code and tools
Change-Id: I0d67e13e9d614bf973a0d4332c6b4dcda4c121a4
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
In other implementations this may be used the same way as qmllint's -I
option, to specify additional import paths.
Change-Id: I447712551ff1fb2a7c379712e938f778b37c514a
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We need more generic filtering capabilities so that it can, for example,
retrieve all the .qml files in a directory, both as qrc paths and local
file paths.
Change-Id: I72a72abc6dd39adb41bcd035f7aa6777e50cb5a5
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
--resource-path directly specifies the resource path for the file being
compiled and overrides any information we might retrieve from the qrc
file. However, we might still find other relevant information in the qrc
file.
Change-Id: I587c6e4e81f792f62a28b035ee97b33d12eeb06d
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
qmlcachegen needs to see the qmltypes file for the current project so
that it can query it for C++ types exposed to QML.
This is only relevant when generating C++ code, which qmlcachegen itself
cannot do. However, in order to write a compatible drop-in, we need it.
Also, hide related ignored options from --help in qmlcachegen.
Change-Id: Id2f1b8b1750351c7de8dfe49e4065ef1b29423b7
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
This signals qmlcachegenplus to not paste the whole JavaScript type
system into each generated file. In turn, user projects need to add a
dependency in order to build against the JS runtime.
qmlcachegen ignores the option.
Change-Id: I0f87dedb969e99e94fbb712b7faa23d84f76dfbe
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
qmlcachegenplus has an argument --direct-calls which instructs it to
call methods of known C++ types directly rather than through the JS
engine or the metaobject. This is faster but requires that all those
types are visible to the generated code. Introduce a way to pass this
option on a per-target base.
Adjust qmlcachegen to ignore the option.
Change-Id: I474e577e4a197f6ca4c8b8e868dfd39983e77041
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We need to re-use them. Also, provide a way to insert AOT compiled
functions into the C++ code.
Change-Id: I7b0d13cb307e8f979745f096a9614f087d135f68
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
This enables us to use the same generateloader.cpp for qmlcachegen and
any replacement that actually produces AOT compiled functions.
Change-Id: I12fe81236e4ef16a627729c644d54b6c171b3860
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We need to be able to generate the loader code from multiple places.
Change-Id: I9e04fd3583b535bc5f7d5fb293cb61309c1e199a
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
For backwards compatibility, a replacement for qmlcachegen will need to
provide the same functionality.
Change-Id: I22664230ea636d384190122223d15819ebee930c
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
The file class names should start with a common prefix.
Change-Id: I5e014c103668a1bc73d359b00a1dda33e5637a79
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Move all the code from tools/shared into src/qmlcompiler and build a
static library from it so that we can re-use it in external tools.
Change-Id: I7c8d8e59063dc7c711f4072f103a01095e6f5997
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
This is needed in a few places outside of declarative, so this change
restores the loc member in DiagnosticMessage and moves
QQmlJS::AST::SourceLocation into common's QQmlJS namespace/directory.
QQmlError is unaffected and retains only line/column.
Amends d4d197d062
Change-Id: Ifb9d344228e3c6e9e26fc4fe112686f9336ea2b2
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Removed dependencies.yaml because we don't use it yet in wip/cmake.
Fixed conflict in qmlcachegen.cpp.
Change-Id: Ie1060c737bee1daa85779903598e5b6d5020d922
Commit 41864db3b6 removed the processing
of .qrc files that removed .qml/etc. files. Since the chaining of
resources remains critical to ensure that the cached compilation units
are loaded, the build system copied the input .qrc file to a new one.
That mere copying caused the build to break when the copied .qrc file
was not in the same directory as the original one, as file paths within
the .qrc file are interpreted as relative to the .qrc file, which was
now in a different location. To fix this, this patch brings back the
"filtering" code that rewrites the paths to the source files in the .qrc
file.
Fixes: QTBUG-78253
Change-Id: Ie1d56f3248e713a964260bc2da37c9374f7b6a36
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
[ChangeLog][QtQml][Important Behavior Changes] Using the Qt Quick
Compiler would exclude the original .qml files from the resource system.
This made it impossible to change the Qt library binary later as the
program binary was tied the to the exact Qt version. In addition
sometimes unrelated files (QTBUG-73669) were removed. For the latter
scenario, retain and skip options were added for the Qt Quick Compiler.
In Qt 5.15 the Qt Quick Compiler does not remove the input files
anymore. All files are retained and the compiler merely adds the more
efficient binary representation to the application.
Task-number: QTBUG-73669
Change-Id: I5a523bfc69d4f48a1451bd880616c82fd73b8d15
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Ulf Hermann
Olivier De Cannière
Joerg Bornemann
Marc Mutz
Ivan Solovev
Maximilian Goldstein
Lucie Gérard
Andrei Golubev
Pavol Markovic
Alexandru Croitor
Simon Hausmann
Qt Forward Merge Bot
Tasuku Suzuki