Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jean-Marc Lasgouttes]

Le 05/12/2022 à 15:54, Kornel Benko a écrit :

I have some work with cmake to do. Have to rearrange the detection of QT and
LYX_STDLIB_DEBUG. No big deal.


OK


But I didn't understand:
Is _GLIBCXX_DEBUG and _GLIBCXX_DEBUG_PEDANTIC now abandoned also for QT5?


This is still there and still named stdlib-debug, but all by default in 
all cases.


The new stdlib-assertions is enabled automatically for development 
releases that use GNU libstdc++ as standard library.


JMarc

--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Kornel Benko]

Am Sun, 4 Dec 2022 19:38:10 +0100
schrieb Jean-Marc Lasgouttes :

> Le 25/11/2022 à 13:00, Kornel Benko a écrit :
> > So, would it be sufficient to use _GLIBCXX_ASSERTIONS in case of QT6 only?
> > Or should we use it on all qt versions?  
> 
> I added in autoconf --enable-stdlib-assertions, that is set 
> automatically in development builds.
> 
> stdlib-debug is now always off and has to be enabled explicitly.
> I could add a warning when this is used together with Qt6 (we do that 
> with hunspell).
> 
> Is this OK with you Kornel?
> 
> JMarc

I have some work with cmake to do. Have to rearrange the detection of QT and
LYX_STDLIB_DEBUG. No big deal.

But I didn't understand:
Is _GLIBCXX_DEBUG and _GLIBCXX_DEBUG_PEDANTIC now abandoned also for QT5?

Kornel


pgp0fCPHp7gGc.pgp
Description: Digitale Signatur von OpenPGP
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Am Sonntag, dem 04.12.2022 um 19:40 +0100 schrieb Jean-Marc Lasgouttes:
> Why do you compile with included-dtl BTW?

Can't remember. It ended up in the build script eventually.

> Maintainer mode should also by the default for development builds.

This as well.

-- 
Jürgen
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jean-Marc Lasgouttes]

Le 19/11/2022 à 20:39, Jürgen Spitzmüller a écrit :

Am Samstag, dem 19.11.2022 um 18:40 +0100 schrieb Jean-Marc Lasgouttes:

But there is no stdlib-debug in this setting, right?


I compile

--with-version-suffix=-dev --enable-maintainer-mode --with-included-
hunspell --with-included-dtl --enable-qt6


Why do you compile with included-dtl BTW?
Maintainer mode should also by the default for development builds.

JMarc

--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jean-Marc Lasgouttes]

Le 25/11/2022 à 13:00, Kornel Benko a écrit :

So, would it be sufficient to use _GLIBCXX_ASSERTIONS in case of QT6 only?
Or should we use it on all qt versions?


I added in autoconf --enable-stdlib-assertions, that is set 
automatically in development builds.


stdlib-debug is now always off and has to be enabled explicitly.
I could add a warning when this is used together with Qt6 (we do that 
with hunspell).


Is this OK with you Kornel?

JMarc
--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Kornel Benko]

Ping

Am Fri, 25 Nov 2022 13:00:01 +0100
schrieb Kornel Benko :

> > > So, maybe the problem is very specific to --enable-stdlib-debug and has 
> > > to do with the way we use external libraries.
> > 
> > It might be time to dump debug iterators and use the less annoying but 
> > less powerful _GLIBCXX_ASSERTIONS, which only does basic index checking.
> > It seems to me that most interesting errors found by stdlib-debug in the 
> > past were in this category.
> > 
> > It was introduced in gcc 6 and the release notes say:
> > "Assertions to check function preconditions can be enabled by defining 
> > the _GLIBCXX_ASSERTIONS macro. The initial set of assertions are a 
> > subset of the checks enabled by the Debug Mode, but without the ABI 
> > changes and changes to algorithmic complexity that are caused by 
> > enabling the full Debug Mode. "
> > 
> > As far as I understand, this avoids all compatibility issues. I put 
> > documentation below; currently we use _GLIBCXX_DEBUG and
> > _GLIBCXX_DEBUG_PEDANTIC.
> > 
> > JMarc
> >   
> 
> So, would it be sufficient to use _GLIBCXX_ASSERTIONS in case of QT6 only?
> Or should we use it on all qt versions?
> 
>   Kornel



pgpOFUe2ZQrMd.pgp
Description: Digitale Signatur von OpenPGP
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Kornel Benko]

Am Wed, 23 Nov 2022 15:48:37 +0100
schrieb Jean-Marc Lasgouttes :

> Le 23/11/2022 à 04:13, Richard Kimberly Heck a écrit :
> > In general, from the LyX bug report, it looks like they're building LyX 
> > but not Qt with debug iterators enabled.
> > 
> > That is a bad idea. libstdc++ debug mode breaks ABI. I don't think we 
> > ever return any QMap in Qt across its library boundary, but if we did, 
> > they couldn't use it.  
> 
> > So, maybe the problem is very specific to --enable-stdlib-debug and has 
> > to do with the way we use external libraries.  
> 
> It might be time to dump debug iterators and use the less annoying but 
> less powerful _GLIBCXX_ASSERTIONS, which only does basic index checking.
> It seems to me that most interesting errors found by stdlib-debug in the 
> past were in this category.
> 
> It was introduced in gcc 6 and the release notes say:
> "Assertions to check function preconditions can be enabled by defining 
> the _GLIBCXX_ASSERTIONS macro. The initial set of assertions are a 
> subset of the checks enabled by the Debug Mode, but without the ABI 
> changes and changes to algorithmic complexity that are caused by 
> enabling the full Debug Mode. "
> 
> As far as I understand, this avoids all compatibility issues. I put 
> documentation below; currently we use _GLIBCXX_DEBUG and
> _GLIBCXX_DEBUG_PEDANTIC.
> 
> JMarc
> 

So, would it be sufficient to use _GLIBCXX_ASSERTIONS in case of QT6 only?
Or should we use it on all qt versions?

Kornel


pgpIpYyEaK_3b.pgp
Description: Digitale Signatur von OpenPGP
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Richard Kimberly Heck]

On 11/23/22 05:29, José Matos wrote:

On Tue, 2022-11-22 at 22:13 -0500, Richard Kimberly Heck wrote:

Here's the program to check:

#define _GLIBCXX_DEBUG 1
#include 

int main()
{
    std::unordered_set::iterator it{};
    it == it;
}

That's what was crashing in GCC.

Are you saying that this program will crash when called?

I do not see this compiling with either gcc 12 or 13.


Yes, that's the program that's mentioned in the bug report at GCC. It 
crashes (in whatever version that was) but should not.


Riki


--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jean-Marc Lasgouttes]

Le 23/11/2022 à 04:13, Richard Kimberly Heck a écrit :
In general, from the LyX bug report, it looks like they're building LyX 
but not Qt with debug iterators enabled.


That is a bad idea. libstdc++ debug mode breaks ABI. I don't think we 
ever return any QMap in Qt across its library boundary, but if we did, 
they couldn't use it.


So, maybe the problem is very specific to --enable-stdlib-debug and has 
to do with the way we use external libraries.


It might be time to dump debug iterators and use the less annoying but 
less powerful _GLIBCXX_ASSERTIONS, which only does basic index checking.
It seems to me that most interesting errors found by stdlib-debug in the 
past were in this category.


It was introduced in gcc 6 and the release notes say:
"Assertions to check function preconditions can be enabled by defining 
the _GLIBCXX_ASSERTIONS macro. The initial set of assertions are a 
subset of the checks enabled by the Debug Mode, but without the ABI 
changes and changes to algorithmic complexity that are caused by 
enabling the full Debug Mode. "


As far as I understand, this avoids all compatibility issues. I put 
documentation below; currently we use _GLIBCXX_DEBUG and

_GLIBCXX_DEBUG_PEDANTIC.

JMarc


_GLIBCXX_ASSERTIONS

Undefined by default. When defined, enables extra error checking in 
the form of precondition assertions, such as bounds checking in strings 
and null pointer checks when dereferencing smart pointers.


_GLIBCXX_DEBUG

Undefined by default. When defined, compiles user code using the 
debug mode. When defined, _GLIBCXX_ASSERTIONS is defined automatically, 
so all the assertions enabled by that macro are also enabled in debug 
mode. [*]


_GLIBCXX_DEBUG_PEDANTIC

Undefined by default. When defined while compiling with the debug 
mode, makes the debug mode extremely picky by making the use of 
libstdc++ extensions and libstdc++-specific behavior into errors.



[*] https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[jspitzm]

Am Dienstag, dem 22.11.2022 um 22:13 -0500 schrieb Richard Kimberly
Heck:
> Here's the program to check:
> 
> #define _GLIBCXX_DEBUG 1
> #include 
> 
> int main()
> {
>    std::unordered_set::iterator it{};
>    it == it;
> }
> 
> That's what was crashing in GCC.

Does not crash for me.

> 
> I did notice, too, though in the Qt report that it says:
> 
> 
> In general, from the LyX bug report, it looks like they're building
> LyX 
> but not Qt with debug iterators enabled.
> 
> That is a bad idea. libstdc++ debug mode breaks ABI. I don't think we
> ever return any QMap in Qt across its library boundary, but if we
> did, they couldn't use it.
> 
> If you want to turn on debug mode globally, you need to rebuild any
> 3rd 
> party dependency with it. Of course there are the containers in the 
> debug namespace, but I haven't found a good solution to use them at 
> large in a project. I've tried to isolate them in a local namespace,
> but 
> people won't use MyNS::vector (which is either std::vector or 
> stddebug::vector), they'll just use std::vector.
> 
> 
> So, maybe the problem is very specific to --enable-stdlib-debug and
> has to do with the way we use external libraries.

I hope this is not the answer, and I don't need to compile qt myself in
order to use stdlib-debug.

Jürgen

> 
> Riki

-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[José Matos]

On Tue, 2022-11-22 at 22:13 -0500, Richard Kimberly Heck wrote:
> Here's the program to check:
> 
> #define _GLIBCXX_DEBUG 1
> #include 
> 
> int main()
> {
>    std::unordered_set::iterator it{};
>    it == it;
> }
> 
> That's what was crashing in GCC.

Are you saying that this program will crash when called?

I do not see this compiling with either gcc 12 or 13. We get a warning
when compiling but that is it:

crash.cpp: In function ‘int main()’:
crash.cpp:7:7: warning: ignoring return value of ‘bool
__gnu_debug::operator==(const
_Safe_iterator,
std::__debug::unordered_set, std::forward_iterator_tag>::_Self&,
const _Safe_iterator,
std::__debug::unordered_set, std::forward_iterator_tag>::_Self&)’,
declared with attribute ‘nodiscard’ [-Wunused-result]
7 |it == it;
  |~~~^
In file included from /usr/include/c++/12/debug/unordered_set:48,
 from /usr/include/c++/12/unordered_set:52,
 from crash.cpp:2:
/usr/include/c++/12/debug/safe_iterator.h:470:7: note: declared here
  470 |   operator==(const _Self& __lhs, const _Self& __rhs)
_GLIBCXX_NOEXCEPT
  |   ^~~~

-- 
José Abílio
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Richard Kimberly Heck]

On 11/21/22 01:18, Jürgen Spitzmüller wrote:

Am Sonntag, dem 20.11.2022 um 18:43 -0500 schrieb Richard Kimberly
Heck:

If I remember correctly, there are little crashing programs in the Qt
and GCC bug reports. Can you try compiling those and see if they
crash?

Not full programs AFAICS.


That at least will help pinpoint where the problem is. I guess I'm
wondering if the problem is still in Qt itself somehow.

I still think it is due to the way we use QFuture(Watcher).


Here's the program to check:

#define _GLIBCXX_DEBUG 1
#include 

int main()
{
  std::unordered_set::iterator it{};
  it == it;
}

That's what was crashing in GCC.

I did notice, too, though in the Qt report that it says:


In general, from the LyX bug report, it looks like they're building LyX 
but not Qt with debug iterators enabled.


That is a bad idea. libstdc++ debug mode breaks ABI. I don't think we 
ever return any QMap in Qt across its library boundary, but if we did, 
they couldn't use it.


If you want to turn on debug mode globally, you need to rebuild any 3rd 
party dependency with it. Of course there are the containers in the 
debug namespace, but I haven't found a good solution to use them at 
large in a project. I've tried to isolate them in a local namespace, but 
people won't use MyNS::vector (which is either std::vector or 
stddebug::vector), they'll just use std::vector.



So, maybe the problem is very specific to --enable-stdlib-debug and has 
to do with the way we use external libraries.


Riki


--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Am Sonntag, dem 20.11.2022 um 18:43 -0500 schrieb Richard Kimberly
Heck:
> If I remember correctly, there are little crashing programs in the Qt
> and GCC bug reports. Can you try compiling those and see if they
> crash? 

Not full programs AFAICS.

> That at least will help pinpoint where the problem is. I guess I'm 
> wondering if the problem is still in Qt itself somehow.

I still think it is due to the way we use QFuture(Watcher).

-- 
Jürgen
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Richard Kimberly Heck]

On 11/20/22 12:30, Jürgen Spitzmüller wrote:

Hi André

Am Donnerstag, dem 17.11.2022 um 12:24 + schrieb André Pönitz:

That's linked from

    https://bugreports.qt.io/browse/QTBUG-108604

now.

That ticket has been closed, but our bug is still there. I see it with
gcc 12.

Any more ideas?


If I remember correctly, there are little crashing programs in the Qt 
and GCC bug reports. Can you try compiling those and see if they crash? 
That at least will help pinpoint where the problem is. I guess I'm 
wondering if the problem is still in Qt itself somehow.


Riki


--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Hi André

Am Donnerstag, dem 17.11.2022 um 12:24 + schrieb André Pönitz:
> That's linked from
> 
>    https://bugreports.qt.io/browse/QTBUG-108604
> 
> now.

That ticket has been closed, but our bug is still there. I see it with
gcc 12.

Any more ideas?

-- 
Jürgen
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Am Samstag, dem 19.11.2022 um 18:40 +0100 schrieb Jean-Marc Lasgouttes:
> But there is no stdlib-debug in this setting, right?

I compile

--with-version-suffix=-dev --enable-maintainer-mode --with-included-
hunspell --with-included-dtl --enable-qt6

I think the crash only occurs when stdlib-debug is involved.

-- 
Jürgen
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jean-Marc Lasgouttes]

Le 19/11/2022 à 17:50, Jürgen Spitzmüller a écrit :

Am Samstag, dem 19.11.2022 um 16:56 +0100 schrieb Jean-Marc Lasgouttes:

../master/configure --with-version-suffix=-master 'CXX=clang++-14
-stdlib=libc++'


This one crashes as well.


But there is no stdlib-debug in this setting, right?

JMarc

--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Am Samstag, dem 19.11.2022 um 16:56 +0100 schrieb Jean-Marc Lasgouttes:
> ../master/configure --with-version-suffix=-master 'CXX=clang++-14 
> -stdlib=libc++'

This one crashes as well.

-- 
Jürgen
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Scott Kostyshak]

On Sat, Nov 19, 2022 at 04:56:04PM +0100, Jean-Marc Lasgouttes wrote:
> Le 19/11/2022 à 01:50, Scott Kostyshak a écrit :
> > On Fri, Nov 18, 2022 at 11:39:49PM +0100, Pavel Sanda wrote:
> > > On Fri, Nov 18, 2022 at 04:23:43PM +0100, Jürgen Spitzmüller wrote:
> > > > Am Donnerstag, dem 17.11.2022 um 19:21 +0100 schrieb Jean-Marc
> > > > Lasgouttes:
> > > > > You're right, I read to fast. So GCC ???11 is ok.
> > > > 
> > > > I am not sure I understand.
> > > 
> > > Andre references to Qt bug, which claims that it is underlying gcc 
> > > problem -
> > > with another link to particular gcc bug. And that one claims to be fixed 
> > > with
> > > milestone gcc 11.
> > > 
> > > > I have gcc 12.2.1 and see the crash.
> > > 
> > > Bummer. We either see different bug or the fix is not enough.
> > 
> > Jürgen, can you compile with Clang to see if you can reproduce?
> > 
> > In case you have never compiled with Clang before, I think it's
> > sufficient to do the following:
> > 
> > export CC="clang"
> > export CXX="clang++"
> 
> Last time I checked, clang++ used GNU stdlibc++. To use clang's libc++
> instead, I do:
> 
> ../master/configure --with-version-suffix=-master 'CXX=clang++-14
> -stdlib=libc++'

Ah, thanks for this correction, JMarc!

Scott


signature.asc
Description: PGP signature
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jean-Marc Lasgouttes]

Le 19/11/2022 à 01:50, Scott Kostyshak a écrit :

On Fri, Nov 18, 2022 at 11:39:49PM +0100, Pavel Sanda wrote:

On Fri, Nov 18, 2022 at 04:23:43PM +0100, Jürgen Spitzmüller wrote:

Am Donnerstag, dem 17.11.2022 um 19:21 +0100 schrieb Jean-Marc
Lasgouttes:

You're right, I read to fast. So GCC ???11 is ok.


I am not sure I understand.


Andre references to Qt bug, which claims that it is underlying gcc problem -
with another link to particular gcc bug. And that one claims to be fixed with
milestone gcc 11.


I have gcc 12.2.1 and see the crash.


Bummer. We either see different bug or the fix is not enough.


Jürgen, can you compile with Clang to see if you can reproduce?

In case you have never compiled with Clang before, I think it's
sufficient to do the following:

export CC="clang"
export CXX="clang++"


Last time I checked, clang++ used GNU stdlibc++. To use clang's libc++ 
instead, I do:


../master/configure --with-version-suffix=-master 'CXX=clang++-14 
-stdlib=libc++'

JMarc
--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Scott Kostyshak]

On Sat, Nov 19, 2022 at 11:48:58AM +0100, Jürgen Spitzmüller wrote:
> Am Freitag, dem 18.11.2022 um 19:50 -0500 schrieb Scott Kostyshak:
> > Jürgen, can you compile with Clang to see if you can reproduce?
> 
> Yes, this also crashes:
> 
> (gdb) bt
> #0  0x760957ec in __pthread_kill_implementation () at
> /lib64/libc.so.6
> #1  0x76042816 in raise () at /lib64/libc.so.6
> #2  0x7602b81c in abort () at /lib64/libc.so.6
> #3  0x762ac141 in std::__throw_bad_exception() () at
> /lib64/libstdc++.so.6
> #4  0x56086e36 in
> __gnu_debug::_Safe_iterator const, QtPrivate::ResultItem> >, std::__debug::map QtPrivate::ResultItem, std::less, std::allocator const, QtPrivate::ResultItem> > >,
> std::forward_iterator_tag>::operator->() const
> (this=this@entry=0x7fffd128)
> at /usr/bin/../lib64/gcc/x86_64-suse-
> linux/12/../../../../include/c++/12/debug/safe_iterator.h:318
> #5  0x56089d59 in QMap QtPrivate::ResultItem>::const_iterator::value() const
> (this=0x7fffd128) at /usr/include/qt6/QtCore/qmap.h:513
> #6  QtPrivate::ResultIteratorBase::pointer()
> const (this=0x7fffd128) at
> /usr/include/qt6/QtCore/qresultstore.h:81
> #7  QtPrivate::ResultIteratorBase::pointer()
> (this=0x7fffd128) at /usr/include/qt6/QtCore/qresultstore.h:74
> #8  QtPrivate::ResultIteratorBase::value()
> (this=0x7fffd128) at /usr/include/qt6/QtCore/qresultstore.h:68
> #9  QFutureInterface::resultReference(int)
> const (this=this@entry=0x57515a80, index=index@entry=0) at
> /usr/include/qt6/QtCore/qfutureinterface.h:373
> #10 0x5606861d in
> QFuture::result void>() const (this=0x57515a80) at
> /usr/include/qt6/QtCore/qfuture.h:308
> #11
> QFutureWatcher::result atus, void>() const (this=) at
> /usr/include/qt6/QtCore/qfuturewatcher.h:113
> #12 lyx::frontend::GuiView::processingThreadFinished()
> (this=0x57701f10) at GuiView.cpp:948
> #13 0x767a6b83 in  () at /lib64/libQt6Core.so.6
> #14 0x768adce8 in QFutureWatcherBase::event(QEvent*) () at
> /lib64/libQt6Core.so.6
> #15 0x777bc543 in QApplicationPrivate::notify_helper(QObject*,
> QEvent*) () at /lib64/libQt6Widgets.so.6
> #16 0x56037036 in
> lyx::frontend::GuiApplication::notify(QObject*, QEvent*) (this=0xa2e,
> receiver=0xa2e, event=0x6) at GuiApplication.cpp:3059
> #17 0x767510f8 in QCoreApplication::notifyInternal2(QObject*,
> QEvent*) () at /lib64/libQt6Core.so.6
> #18 0x767512e0 in
> QCoreApplicationPrivate::sendPostedEvents(QObject*, int, QThreadData*)
> () at /lib64/libQt6Core.so.6
> #19 0x76967613 in  () at /lib64/libQt6Core.so.6
> #20 0x75ddba90 in g_main_context_dispatch () at /lib64/libglib-
> 2.0.so.0
> #21 0x75ddbe48 in  () at /lib64/libglib-2.0.so.0
> #22 0x75ddbedc in g_main_context_iteration () at
> /lib64/libglib-2.0.so.0
> #23 0x76965b0c in
> QEventDispatcherGlib::processEvents(QFlags g>) () at /lib64/libQt6Core.so.6
> #24 0x7675b98b in
> QEventLoop::exec(QFlags) () at
> /lib64/libQt6Core.so.6
> #25 0x76754538 in QCoreApplication::exec() () at
> /lib64/libQt6Core.so.6
> #26 0x55c0292d in lyx::LyX::exec(int&, char**)
> (this=0x7fffd7e0, argc=, argv=) at
> LyX.cpp:397
> #27 0x55a36ff9 in main(int, char**) (argc=,
> argv=0x6) at main.cpp:55

Thanks for checking, Jürgen. Things are not as simple as we thought, it seems.

Scott


signature.asc
Description: PGP signature
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Am Freitag, dem 18.11.2022 um 19:50 -0500 schrieb Scott Kostyshak:
> Jürgen, can you compile with Clang to see if you can reproduce?

Yes, this also crashes:

(gdb) bt
#0  0x760957ec in __pthread_kill_implementation () at
/lib64/libc.so.6
#1  0x76042816 in raise () at /lib64/libc.so.6
#2  0x7602b81c in abort () at /lib64/libc.so.6
#3  0x762ac141 in std::__throw_bad_exception() () at
/lib64/libstdc++.so.6
#4  0x56086e36 in
__gnu_debug::_Safe_iterator >, std::__debug::map, std::allocator > >,
std::forward_iterator_tag>::operator->() const
(this=this@entry=0x7fffd128)
at /usr/bin/../lib64/gcc/x86_64-suse-
linux/12/../../../../include/c++/12/debug/safe_iterator.h:318
#5  0x56089d59 in QMap::const_iterator::value() const
(this=0x7fffd128) at /usr/include/qt6/QtCore/qmap.h:513
#6  QtPrivate::ResultIteratorBase::pointer()
const (this=0x7fffd128) at
/usr/include/qt6/QtCore/qresultstore.h:81
#7  QtPrivate::ResultIteratorBase::pointer()
(this=0x7fffd128) at /usr/include/qt6/QtCore/qresultstore.h:74
#8  QtPrivate::ResultIteratorBase::value()
(this=0x7fffd128) at /usr/include/qt6/QtCore/qresultstore.h:68
#9  QFutureInterface::resultReference(int)
const (this=this@entry=0x57515a80, index=index@entry=0) at
/usr/include/qt6/QtCore/qfutureinterface.h:373
#10 0x5606861d in
QFuture::result() const (this=0x57515a80) at
/usr/include/qt6/QtCore/qfuture.h:308
#11
QFutureWatcher::result() const (this=) at
/usr/include/qt6/QtCore/qfuturewatcher.h:113
#12 lyx::frontend::GuiView::processingThreadFinished()
(this=0x57701f10) at GuiView.cpp:948
#13 0x767a6b83 in  () at /lib64/libQt6Core.so.6
#14 0x768adce8 in QFutureWatcherBase::event(QEvent*) () at
/lib64/libQt6Core.so.6
#15 0x777bc543 in QApplicationPrivate::notify_helper(QObject*,
QEvent*) () at /lib64/libQt6Widgets.so.6
#16 0x56037036 in
lyx::frontend::GuiApplication::notify(QObject*, QEvent*) (this=0xa2e,
receiver=0xa2e, event=0x6) at GuiApplication.cpp:3059
#17 0x767510f8 in QCoreApplication::notifyInternal2(QObject*,
QEvent*) () at /lib64/libQt6Core.so.6
#18 0x767512e0 in
QCoreApplicationPrivate::sendPostedEvents(QObject*, int, QThreadData*)
() at /lib64/libQt6Core.so.6
#19 0x76967613 in  () at /lib64/libQt6Core.so.6
#20 0x75ddba90 in g_main_context_dispatch () at /lib64/libglib-
2.0.so.0
#21 0x75ddbe48 in  () at /lib64/libglib-2.0.so.0
#22 0x75ddbedc in g_main_context_iteration () at
/lib64/libglib-2.0.so.0
#23 0x76965b0c in
QEventDispatcherGlib::processEvents(QFlags) () at /lib64/libQt6Core.so.6
#24 0x7675b98b in
QEventLoop::exec(QFlags) () at
/lib64/libQt6Core.so.6
#25 0x76754538 in QCoreApplication::exec() () at
/lib64/libQt6Core.so.6
#26 0x55c0292d in lyx::LyX::exec(int&, char**)
(this=0x7fffd7e0, argc=, argv=) at
LyX.cpp:397
#27 0x55a36ff9 in main(int, char**) (argc=,
argv=0x6) at main.cpp:55


-- 
Jürgen


signature.asc
Description: This is a digitally signed message part
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Am Freitag, dem 18.11.2022 um 23:39 +0100 schrieb Pavel Sanda:
> Bummer. We either see different bug or the fix is not enough.

Maybe. All I can say is that I still see the very bug I reported.

-- 
Jürgen
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Scott Kostyshak]

On Fri, Nov 18, 2022 at 11:39:49PM +0100, Pavel Sanda wrote:
> On Fri, Nov 18, 2022 at 04:23:43PM +0100, Jürgen Spitzmüller wrote:
> > Am Donnerstag, dem 17.11.2022 um 19:21 +0100 schrieb Jean-Marc
> > Lasgouttes:
> > > You're right, I read to fast. So GCC ???11 is ok.
> > 
> > I am not sure I understand.
> 
> Andre references to Qt bug, which claims that it is underlying gcc problem -
> with another link to particular gcc bug. And that one claims to be fixed with
> milestone gcc 11.
> 
> > I have gcc 12.2.1 and see the crash.
> 
> Bummer. We either see different bug or the fix is not enough.

Jürgen, can you compile with Clang to see if you can reproduce?

In case you have never compiled with Clang before, I think it's
sufficient to do the following:

export CC="clang"
export CXX="clang++"

Scott


signature.asc
Description: PGP signature
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Pavel Sanda]

On Fri, Nov 18, 2022 at 04:23:43PM +0100, Jürgen Spitzmüller wrote:
> Am Donnerstag, dem 17.11.2022 um 19:21 +0100 schrieb Jean-Marc
> Lasgouttes:
> > You're right, I read to fast. So GCC ???11 is ok.
> 
> I am not sure I understand.

Andre references to Qt bug, which claims that it is underlying gcc problem -
with another link to particular gcc bug. And that one claims to be fixed with
milestone gcc 11.

> I have gcc 12.2.1 and see the crash.

Bummer. We either see different bug or the fix is not enough.

Pavel
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Am Donnerstag, dem 17.11.2022 um 19:21 +0100 schrieb Jean-Marc
Lasgouttes:
> You're right, I read to fast. So GCC ≥11 is ok.

I am not sure I understand. I have gcc 12.2.1 and see the crash.

-- 
Jürgen
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jean-Marc Lasgouttes]

You're right, I read to fast. So GCC ≥11 is ok.

We could also drop stdlib-debug and use the less intrusive assertion define 
(sorry I can't look it up now).

JMarc 

Le 17 novembre 2022 18:26:48 GMT+01:00, Richard Kimberly Heck 
 a écrit :
>On 11/17/22 12:18, Jean-Marc Lasgouttes wrote:
>> Le 17/11/2022 à 16:43, Scott Kostyshak a écrit :
>>> Thanks so much, Andre'! I really appreciate that!
>>> 
>>> It is very helpful to know that it is a GCC bug. Now we can make a plan
>>> moving forward for the release regarding this issue.
>> 
>> I read it as a Qt bug: "we're comparing singular (= default-constructed) 
>> std:map const_iterator'ss, which is UB [undefined behavior?]."
>
>But they change their mind about that. The specification says this should be 
>ok. In the end, it was fixed in GCC.
>
>Riki
>
>
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Richard Kimberly Heck]

On 11/17/22 12:18, Jean-Marc Lasgouttes wrote:

Le 17/11/2022 à 16:43, Scott Kostyshak a écrit :

Thanks so much, Andre'! I really appreciate that!

It is very helpful to know that it is a GCC bug. Now we can make a plan
moving forward for the release regarding this issue.


I read it as a Qt bug: "we're comparing singular (= 
default-constructed) std:map const_iterator'ss, which is UB [undefined 
behavior?]."


But they change their mind about that. The specification says this 
should be ok. In the end, it was fixed in GCC.


Riki


--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Pavel Sanda]

On Thu, Nov 17, 2022 at 10:59:59AM -0500, Richard Kimberly Heck wrote:
> If I'm reading the GCC stuff correctly, the bug is in the C++ libraries, so
> the bug is not really in the LyX binaries at all, or even in the Qt ones. We
> see it with Qt 6 because they use a construction there that triggers the
> bug.

Yep. It seems that poor man's solution is to enforce --disable-stdlib-debug
on configurations with Qt6 & gcc < 11.

Pavel
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jean-Marc Lasgouttes]

Le 17/11/2022 à 16:43, Scott Kostyshak a écrit :

Thanks so much, Andre'! I really appreciate that!

It is very helpful to know that it is a GCC bug. Now we can make a plan
moving forward for the release regarding this issue.


I read it as a Qt bug: "we're comparing singular (= default-constructed) 
std:map const_iterator'ss, which is UB [undefined behavior?]."


JMarc

PS: nice to read from you André, and to see that you are still there 
when we need you :)

--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Richard Kimberly Heck]

On 11/17/22 10:48, Scott Kostyshak wrote:

On Thu, Nov 17, 2022 at 10:43:40AM -0500, Scott Kostyshak wrote:

On Thu, Nov 17, 2022 at 12:24:31PM +, André Pönitz wrote:

That's linked from

https://bugreports.qt.io/browse/QTBUG-108604


It is very helpful to know that it is a GCC bug. Now we can make a plan
moving forward for the release regarding this issue.

Does this mean that if LyX is compiled with a compiler without the bug
(e.g., Clang or perhaps a different GCC without the bug), then the bug
will not be present in the LyX binary? Or does it mean that the Qt
library itself needs to have been compiled with a compiler without that
bug?


If I'm reading the GCC stuff correctly, the bug is in the C++ libraries, 
so the bug is not really in the LyX binaries at all, or even in the Qt 
ones. We see it with Qt 6 because they use a construction there that 
triggers the bug.


Riki


--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Scott Kostyshak]

On Thu, Nov 17, 2022 at 10:43:40AM -0500, Scott Kostyshak wrote:
> On Thu, Nov 17, 2022 at 12:24:31PM +, André Pönitz wrote:

> > That's linked from
> > 
> >https://bugreports.qt.io/browse/QTBUG-108604
> > 
> It is very helpful to know that it is a GCC bug. Now we can make a plan
> moving forward for the release regarding this issue.

Does this mean that if LyX is compiled with a compiler without the bug
(e.g., Clang or perhaps a different GCC without the bug), then the bug
will not be present in the LyX binary? Or does it mean that the Qt
library itself needs to have been compiled with a compiler without that
bug?

Scott


signature.asc
Description: PGP signature
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Scott Kostyshak]

On Thu, Nov 17, 2022 at 12:24:31PM +, André Pönitz wrote:
> Hi.
> 
> On Fri, Nov 11, 2022 at 03:16:05PM -0500, Scott Kostyshak wrote:
> > The following issue seems pretty bad:
> >
> >   https://www.lyx.org/trac/ticket/12215
> 
> That's linked from
> 
>https://bugreports.qt.io/browse/QTBUG-108604
> 
> now.

Thanks so much, Andre'! I really appreciate that!

It is very helpful to know that it is a GCC bug. Now we can make a plan
moving forward for the release regarding this issue.

Thanks to you and your team,

Scott


signature.asc
Description: PGP signature
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[André Pönitz]

Hi.

On Fri, Nov 11, 2022 at 03:16:05PM -0500, Scott Kostyshak wrote:
> The following issue seems pretty bad:
>
>   https://www.lyx.org/trac/ticket/12215

That's linked from

   https://bugreports.qt.io/browse/QTBUG-108604

now.

Best regards,
Andre'
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Scott Kostyshak]

On Fri, Nov 11, 2022 at 03:16:05PM -0500, Scott Kostyshak wrote:
> The following issue seems pretty bad:
> 
>   https://www.lyx.org/trac/ticket/12215
> 
> Is there someone who is brave enough with time to attempt a fix?
> 
> If no one fixes it in time, shall we postpone 2.4.0 or proceed with
> 2.4.0 and state that Qt6 is not officially supported because of #12215?
> I suppose an alternative would be to disable asynch on autosave and
> export/view when compiling with Qt6?
> 
> Scott

Hi André (I am CC'ing the lyx-devel list),

There is a Qt6-specific issue that we are stuck on. I am sorry to bother
you, but we are a bit desperate since it is essentially a release
blocker and yet we do not have an idea for how to fix it. Do you by
chance have time to take a look? The ticket is here:

  https://www.lyx.org/trac/ticket/12215

Thanks,
Scott
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Am Montag, dem 14.11.2022 um 13:47 -0500 schrieb Scott Kostyshak:
> I don't suppose we could ask André for a favor to look into it? I do
> not know him, but I am happy to ask if others think it is not a bad
> idea.

Yes, it's worth a try.

-- 
Jürgen
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Cor Blom]

Op 14-11-2022 om 13:46 schreef Thibaut Cuvelier:
On Mon, 14 Nov 2022, 11:45 Pavel Sanda, > wrote:


On Fri, Nov 11, 2022 at 03:16:05PM -0500, Scott Kostyshak wrote:
 > If no one fixes it in time, shall we postpone 2.4.0

No, I wouldn't do that.

 > or proceed with 2.4.0 and state that Qt6 is not officially
supported because of #12215?

I see three options:
1) we don't support QT6 and break during ./configure if detected
2) we allow Qt6 with --enable-stdlib-debug
3) we allow Qt6 but forcing single thread only.

I don't have clear winner between these three.


I think that not supporting Qt 6 is the worst option, especially as it 
may create problems with package managers if they want to switch to Qt 6 
for any reason. Requiring a parameter to be set at configure time to 
allow Qt 6 acknowledging the potential problems would be acceptable, in 
my opinion.


As maintainer of the openSUSE packages I would like to be able to build 
with Qt6 when KDE Plasma 6 will be released.


https://community.kde.org/Schedules/Plasma_6

Cor



--
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Scott Kostyshak]

On Sat, Nov 12, 2022 at 06:28:44AM +0100, Jürgen Spitzmüller wrote:
> Am Freitag, dem 11.11.2022 um 21:32 +0100 schrieb Thibaut Cuvelier:
> > > The following issue seems pretty bad:
> > > 
> > >   https://www.lyx.org/trac/ticket/12215
> > > 
> > > Is there someone who is brave enough with time to attempt a fix?
> 
> I am afraid it is beyond me (I tried to delve into it but gave up)

I don't suppose we could ask André for a favor to look into it? I do not
know him, but I am happy to ask if others think it is not a bad idea.

Scott
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Thibaut Cuvelier]

On Mon, 14 Nov 2022, 11:45 Pavel Sanda,  wrote:

> On Fri, Nov 11, 2022 at 03:16:05PM -0500, Scott Kostyshak wrote:
> > If no one fixes it in time, shall we postpone 2.4.0
>
> No, I wouldn't do that.
>
> > or proceed with 2.4.0 and state that Qt6 is not officially supported
> because of #12215?
>
> I see three options:
> 1) we don't support QT6 and break during ./configure if detected
> 2) we allow Qt6 with --enable-stdlib-debug
> 3) we allow Qt6 but forcing single thread only.
>
> I don't have clear winner between these three.
>

I think that not supporting Qt 6 is the worst option, especially as it may
create problems with package managers if they want to switch to Qt 6 for
any reason. Requiring a parameter to be set at configure time to allow Qt 6
acknowledging the potential problems would be acceptable, in my opinion.

>
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Pavel Sanda]

On Fri, Nov 11, 2022 at 03:16:05PM -0500, Scott Kostyshak wrote:
> If no one fixes it in time, shall we postpone 2.4.0

No, I wouldn't do that.

> or proceed with 2.4.0 and state that Qt6 is not officially supported because 
> of #12215?

I see three options:
1) we don't support QT6 and break during ./configure if detected
2) we allow Qt6 with --enable-stdlib-debug
3) we allow Qt6 but forcing single thread only.

I don't have clear winner between these three.

Pavel
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Jürgen Spitzmüller]

Am Freitag, dem 11.11.2022 um 21:32 +0100 schrieb Thibaut Cuvelier:
> > The following issue seems pretty bad:
> > 
> >   https://www.lyx.org/trac/ticket/12215
> > 
> > Is there someone who is brave enough with time to attempt a fix?

I am afraid it is beyond me (I tried to delve into it but gave up)

> 
> Is it possible that this issue is a bug in Qt itself? For instance,
> it looks a bit like https://bugreports.qt.io/browse/QTBUG-6799.

I don't think so (this bug affects Qt4 and 5, not Qt6). I really think
it has to do with the QFutureInterface/QPromise changes in Qt6:
https://doc.qt.io/qt-6/qtcore-changes-qt6.html#qfuture-and-related-classes

-- 
Jürgen
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel

Re: 2.4.0 plan for #12215 [LyX crashes with async processes (Qt6 only)] ?

[Thibaut Cuvelier]

On Fri, 11 Nov 2022 at 21:16, Scott Kostyshak  wrote:

> The following issue seems pretty bad:
>
>   https://www.lyx.org/trac/ticket/12215
>
> Is there someone who is brave enough with time to attempt a fix?
>
> If no one fixes it in time, shall we postpone 2.4.0 or proceed with
> 2.4.0 and state that Qt6 is not officially supported because of #12215?
> I suppose an alternative would be to disable asynch on autosave and
> export/view when compiling with Qt6?
>

Is it possible that this issue is a bug in Qt itself? For instance, it
looks a bit like https://bugreports.qt.io/browse/QTBUG-6799.
-- 
lyx-devel mailing list
lyx-devel@lists.lyx.org
http://lists.lyx.org/mailman/listinfo/lyx-devel