Gpg4win 3.0.3 GpgOL crashes with old mails already verified by GPG4win 2.x
Closed, DuplicatePublic

Description

At work we have Outlook 2010 32-bit, currently version 14.0.7194.5000, with Exchange server.

When I received encrypted mails, I used Thunderbird instead, because it simply didn't work back at Gpg4win 2.x. Nevertheless those mails are still stored in some folder in Outlook and have been "touched" by the old GpgOL. I don't know, if untouched mails would also be a problem.

While current mail flow works well, GpgOL crashes when trying to view those old mails.

My PC is running Windows 7 x64.

JJworx created this task.Mar 23 2018, 8:42 AM
aheinecke claimed this task.
aheinecke triaged this task as High priority.
aheinecke added a subscriber: aheinecke.

Thanks for your report. Sadly I cannot reproduce this, I went back in my archives and even mails from 2015 / touched by Gpg4win 2.x work without a problem.

Until we have analyzed this further I give this high priority anyway as we want to fix any crash in GpgOL.

Is the crash reproduceable for you? Does it happen every time on some mail?

Can you please give us a debug log of such a crash:

  • Close Outlook
  • Open regedit.exe and
  • set HKEY_CURRENT_USER\Software\GNU\GpgOL enableDebug to either 1 or 1665
  • set HKEY_CURRENT_USER\Software\GNU\GpgOL logFile to some writeable file (e.g. c:\users\<youruser>\gpgol.txt )
  • Open Outlook again and reproduce the crash.

The value of enableDebug -> If set to 1 it logs less and should not leak any private data. If set to 1665 the log will contain the plaintext of decrypted messages, so only use that if you have a test mail or some unconfidential mail.

If you are not comfortable attaching the log here you can send it to me privately to aheinecke@intevation.de (PGP Key 94A5C9A03C2FE5CA3B095D8E1FDF723CF462B6B1 )

Thank you for your help.

I answered by mail in this fashion:

I'm surprised the error is not reproducable any more, even though it was so before. Perhaps I had the problem not with Gpg4win 3.0.3, but with 3.1.0 beta, which I had installed prior the same day.

Is the error reproducable for you under Gpg4win 3.1.0? If not, I'll try installing it again.

So, this is tested with 3.1.0 beta 38, reproducable crash

Funny thing, it worked for some time, now it's reproducably crashing again. This might be the better log file.

Apologies for not replying to your mail directly. I've marked it in my INBOX to do the test with 3.0.3 first but have not gotten around to it.

Thank you for testing anyway and providing the logs.

Looking at your logs, you are verifying S/MIME signed mails with a large attachment. I've tried it here ~10 times in repeat with such a mail and 2MB attachment and I did not encounter a crash. So at this moment I don't think that this is the problem.

Funny thing, it worked for some time, now it's reproducably crashing again. This might be the better log file.

Sounds a bit like an unitialized memory access. They often show such a pattern.

I think that the crash is in the get_ultimate_keys function somehow.

In your first log I see:

07:09:31/7012/parsecontroller.cpp:get_ultimate_keys: Starting keylisting.

That is the last line logged from Thread 7012. The parser thread is still running and writing output but I doubt that this is the problem. Because in the second log from you

08:52:43/6900/parsecontroller.cpp:get_ultimate_keys: Starting keylisting.

Is the last line.

I'll review the code now to check if I see something.

Mr. Heinecke, to make sure, please note that despite the thread title these crashes happened with 3.1.0. beta 38. It would be sad if you do all of the tests and checks with 3.0.3

I tried to reproduce this again, using S/MIME Mails, installing gpg4win 2.x etc. It did not crash for me :-/

Is it always the same mail that is crashing? Maybe there is something special about the mail?

With rO857afbf1a2e5 I've added some additional safety checks and debug output in the area of get_ultimate_keys. I don't really see what could cause a crash there but maybe this will help us to further home in on the bug.

So if you can still reproduce the crash with gpg4win-3.1.0-beta48 please add a log again. It should tell me a bit more.

This was likely one form of T4111

Let's reopen this if we still see S/MIME crashes after 3.1.5 is released.