Problems with [Q]GpgME and Qt6

rjmx

(Crossposting to gnupg-users mailing list)

I have an app that works fine in Qt5. I use QGpgME to encrypt/decrypt messages and handle key usage. I'm trying to upgrade the app to Qt6, and, naturally, expected problems, but this one beats me.

Here, for example, is the code I use to retrieve my secret key list:

QGpgME::KeyListJob *job = QGpgME::openpgp()->keyListJob(false, false, false);

// Prepare result vector
std::vector<GpgME::Key> keys;

// Execute it synchronously
GpgME::KeyListResult result = job->exec(QStringList() <<
                                        QStringLiteral(""),  // No pattern: get all
                                        true, keys);        // Only secret keys
 delete job;

Now, this is straight out of the "t-keylist.cpp" test file that comes with the GpgME source. It works perfectly well under Qt5. but under Qt6, it segfaults at the "job->exec" line.

Anyone have a clue why it crashes under Qt6?

Thanks,

--
Ron Murray rjmx@rjmx.net
PGP Fingerprint: 4D99 70E3 2317 334B 141E 7B63 12F7 E865 B5E2 E761

JonB

@rjmx said in Problems with [Q]GpgME and Qt6:

it segfaults at the "job->exec" line.

Where does your code verify the return result from job = QGpgME::openpgp()->keyListJob(false, false, false); for nullptr?
Look at the stack trace in debugger when it segfaults. Are you saying it does on the job->exec(...) call or somewhere inside the exec() function?

rjmx

@JonB Thanks for the reply.
The code doesn't check for a nullptr reply to the keyListJob() call (although I agree that it should), but I'd checked (stepping through the code) that the result was non-zero. Checking now, it comes back with a value of "job @0x555555dbf5d0", which seems reasonable.

I'd looked at the stack trace before, but it didn't seem to tell me much. This time, it gets

1   QString::toUtf8_helper(QString const&)                                            0x7ffff5b63496 
2   ??                                                                                0x7ffff7ef6325 
3   ??                                                                                0x7ffff7eb9a60 
4   ??                                                                                0x7ffff7eba58f 
5   ??                                                                                0x7ffff7ebb67a 
6   Cryptography::getKeyNames                               cryptography.cpp      53  0x55555557a082

Cryptography::getKeyNames() is the function I wrote that contains the code in question. It crashed immediately upon executing the job->exec() call.

Now I guess I need to figure out why it crashes on the QString::toUtf8_helper() call.

.....Ron

rjmx

@rjmx @JonB
Further to this, this happens with all three occasions in this code, where I call QGpgME (finding secret keys, encryption and decryption), but the failing functions differ.

Here's what happens with the decrypting function. The relevant code is

// Create the job
   QGpgME::DecryptJob *decryptjob = QGpgME::openpgp()->decryptJob();
   // Execute it synchronously
   // decryptjob->exec: const QByteArray &cipherText, QByteArray &plainText
   GpgME::DecryptionResult decryptresult;
   decryptresult = decryptjob->exec(ciphertext, plaintext);

and the stack trace becomes


For the encryption task, the 
1   QByteArray::operator=(QByteArray const&)                                          0x7ffff5b33540
2   ??                                                                                0x7ffff7e9b10d
3   Cryptography::decrypt                                      cryptography.cpp   203 0x55555557b1b2

For the encryption task, the code is

// encryptJob(/*ASCII Armor */false, /* Textmode */ false);
   // If 'recipient' is entry, GpgME does password encryption
   QGpgME::EncryptJob *encryptjob = QGpgME::openpgp()->encryptJob(true, true);

   QByteArray ciphertext, plaintext;
   plaintext = content.toUtf8();
   GpgME::EncryptionResult encryptresult = encryptjob->exec(recipient, plaintext, false, ciphertext);

and the stack trace is

1   QString::toUtf8_helper(QString const&)                                            0x7ffff5b63496
2   ??                                                                                0x7ffff7ef6325
3   ??                                                                                0x7ffff7eb9a60
4   ??                                                                                0x7ffff7eba58f
5   ??                                                                                0x7ffff7ebb67a
6   Cryptography::encrypt                                      cryptography.cpp   125 0x55555557a86c

Looks like there's a problem handling text, but not in the same function.

rjmx

@rjmx @JonB
Just tried it again on the keyListJob function, got the same crash on the same function (QString::toUtf8_helper(QString const&)), but this time got into the disassembler crash disassemble.png

Christian Ehrlicher

My guess is that you mix debug and release libraries which you should not do when using msvc.

rjmx

@Christian-Ehrlicher Perhaps I should have said this earlier, but I’m doing this on Linux with gcc. All I can say is that I’ve changed nothing in the switch from Qt5 to Qt6 other than changing the kit (and fixing issues caused by a change in the calling arguments for QCalendarWidget::paintCell()).
Tomorrow, I’ll delve into the source code for QGpgME and see if I can see anything obvious.

Christian Ehrlicher

Ok, then are you sure QGpgME is compiled and linked against Qt6? Check with ldd.

rjmx

@Christian-Ehrlicher
Good point. I'd checked with the Debian maintainers of gpgme, and they replied they were Qt6-compatible (at least, that was my understanding). Taking your advice to try ldd, however, I get

ron:~/$ ldd /usr/lib/x86_64-linux-gnu/libqgpgme.so.15
..
..
..
	libQt5Core.so.5 => /lib/x86_64-linux-gnu/libQt5Core.so.5 (0x00007f8bcf600000)

So, Qt5Core but no mention of Qt6Core? That doesn't look good.

I'll look into rebuilding it myself, I think.