Details
-
Bug
-
Resolution: Invalid
-
Not Evaluated
-
None
-
6.7.2
-
None
-
all platforms, Qt's internal PCRE2 default to JIT mode
-
All
Description
If you read the PCRE2 docs, the use of the JIT will make it unsafe in multithreaded environments without very special handling which is typically never employed.
The problem is QRegularExpression uses PCRE2 and enables the JIT which makes the use of QRegularExpression in routines that are invoked by QtConcurrent a big problem causing random crashes in the ~QExplicitlySharedDataPointer as follows:
Target 0: (Sigil) stopped. (lldb) bt * thread #48, name = 'Thread (pooled)', stop reason = EXC_BAD_ACCESS (code=1, address=0x228095ff8) * frame #0: 0x00000001078d4940 QtCore`sljit_free_exec + 240 frame #1: 0x00000001078e8bb1 QtCore`_pcre2_jit_free_16 + 129 frame #2: 0x00000001078c441f QtCore`pcre2_code_free_16 + 31 frame #3: 0x000000010772f424 QtCore`QExplicitlySharedDataPointer<QRegularExpressionPrivate>::~QExplicitlySharedDataPointer() + 36 frame #4: 0x00000001003c29b5 Sigil`GumboInterface::update_style_urls(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&) + 1701 frame #5: 0x00000001003c3576 Sigil`GumboInterface::build_attributes(GumboAttribute*, bool, bool, bool) + 1366 frame #6: 0x00000001003b8f3d Sigil`GumboInterface::serialize(GumboInternalNode*, GumboInterface::UpdateTypes) + 829 frame #7: 0x00000001003bd7f2 Sigil`GumboInterface::serialize_contents(GumboInternalNode*, GumboInterface::UpdateTypes) + 402 frame #8: 0x00000001003b928f Sigil`GumboInterface::serialize(GumboInternalNode*, GumboInterface::UpdateTypes) + 1679 frame #9: 0x00000001003bd7f2 Sigil`GumboInterface::serialize_contents(GumboInternalNode*, GumboInterface::UpdateTypes) + 402 frame #10: 0x00000001003b928f Sigil`GumboInterface::serialize(GumboInternalNode*, GumboInterface::UpdateTypes) + 1679 frame #11: 0x00000001003bd7f2 Sigil`GumboInterface::serialize_contents(GumboInternalNode*, GumboInterface::UpdateTypes) + 402 frame #12: 0x00000001003b928f Sigil`GumboInterface::serialize(GumboInternalNode*, GumboInterface::UpdateTypes) + 1679 frame #13: 0x00000001003bd7f2 Sigil`GumboInterface::serialize_contents(GumboInternalNode*, GumboInterface::UpdateTypes) + 402 frame #14: 0x00000001003b928f Sigil`GumboInterface::serialize(GumboInternalNode*, GumboInterface::UpdateTypes) + 1679 frame #15: 0x00000001003bd7f2 Sigil`GumboInterface::serialize_contents(GumboInternalNode*, GumboInterface::UpdateTypes) + 402 frame #16: 0x00000001003b928f Sigil`GumboInterface::serialize(GumboInternalNode*, GumboInterface::UpdateTypes) + 1679 frame #17: 0x00000001003bd7f2 Sigil`GumboInterface::serialize_contents(GumboInternalNode*, GumboInterface::UpdateTypes) + 402 frame #18: 0x00000001003b8d31 Sigil`GumboInterface::serialize(GumboInternalNode*, GumboInterface::UpdateTypes) + 305 frame #19: 0x00000001003bca35 Sigil`GumboInterface::perform_style_updates(QString const&, QString const&) + 309 frame #20: 0x0000000100058bba Sigil`PerformHTMLUpdates::operator()() + 234 frame #21: 0x000000010006d4df Sigil`UniversalUpdates::UpdateOneHTMLFile(HTMLResource*, QHash<QString, QString> const&, QHash<QString, QString> const&) + 239 frame #22: 0x00000001000701fa Sigil`QtConcurrent::MappedEachKernel<QList<HTMLResource*>::const_iterator, std::__1::__bind<QString (&)(HTMLResource*, QHash<QString, QString> const&, QHash<QString, QString> const&), std::__1::placeholders::__ph<1> const&, QHash<QString, QString>&, QHash<QString, QString>&> >::runIteration(QList<HTMLResource*>::const_iterator, int, QString*) + 42 frame #23: 0x0000000100070292 Sigil`QtConcurrent::MappedEachKernel<QList<HTMLResource*>::const_iterator, std::__1::__bind<QList<std::__1::pair<QString, QString> > (&)(HTMLResource*, QHash<QString, CSSInfo*> const&), std::__1::placeholders::__ph<1> const&, QHash<QString, CSSInfo*>&> >::runIterations(QList<HTMLResource*>::const_iterator, int, int, QList<std::__1::pair<QString, QString> >*) + 66 frame #24: 0x00000001000705a7 Sigil`QtConcurrent::IterateKernel<QList<HTMLResource*>::const_iterator, QString>::forThreadFunction() + 359 frame #25: 0x0000000100947996 QtConcurrent`QtConcurrent::ThreadEngineBase::run() + 134 frame #26: 0x00000001076fb741 QtCore`QThreadPoolThread::run() + 225 frame #27: 0x00000001076f39c9 QtCore`QThreadPrivate::start(void*) + 329 frame #28: 0x00007ff8026851d3 libsystem_pthread.dylib`_pthread_start + 125
If I force the use of:
QT_ENABLE_REGEXP_JIT=0
everything works.
But nowhere does it say that QRegularExpression is thread unsafe and that its code should not be used with Qt:Concurrent.
If you plan to default to using the PCRE2 JIT withing QRegularExpression, you should properly handle the its use to make it thread safe.
See the PCRE2 docs here:
https://www.pcre.org/current/doc/html/pcre2jit.html#SEC7
And especially this part:
In a multithread application, if you do not specify a JIT stack, or if you assign or pass back NULL from a callback, that is thread-safe, because each thread has its own machine stack. However, if you assign or pass back a non-NULL JIT stack, this must be a different stack for each thread so that the application is thread-safe. Strictly speaking, even more is allowed. You can assign the same non-NULL stack to a match context that is used by any number of patterns, as long as they are not used for matching by multiple threads at the same time. For example, you could use the same stack in all compiled patterns, with a global mutex in the callback to wait until the stack is available for use. However, this is an inefficient solution, and not recommended. This is a suggestion for how a multithreaded program that needs to set up non-default JIT stacks might operate: During thread initialization thread_local_var = pcre2_jit_stack_create(...) During thread exit pcre2_jit_stack_free(thread_local_var) Use a one-line callback function return thread_local_var
Please fix QRegularExpression to be thread safe.