Re: Authentication segfault with Dovecot 2.3.13

2021-01-14 Thread Aki Tuomi
Hi! This has been fixed with https://github.com/dovecot/core/commit/225e2e040c5725753eb80a725d3d36878ba9a243.patch Aki > On 11/01/2021 13:24 Harald Leithner wrote: > > > Hi, > > just a confirmation after removing 'nodelay' we didn't notice any > further segfaults. > > thanks > > Harald

Re: Authentication segfault with Dovecot 2.3.13

2021-01-11 Thread Harald Leithner
Hi, just a confirmation after removing 'nodelay' we didn't notice any further segfaults. thanks Harald Am 07.01.2021 um 09:05 schrieb Harald Leithner: Hi, 1. legacy customer requirement... 2. we will try to remove nodelay thanks Am 07.01.2021 um 00:05 schrieb Josef 'Jeff' Sipek: On Wed,

Re: Authentication segfault with Dovecot 2.3.13

2021-01-07 Thread Harald Leithner
Hi, 1. legacy customer requirement... 2. we will try to remove nodelay thanks Am 07.01.2021 um 00:05 schrieb Josef 'Jeff' Sipek: On Wed, Jan 06, 2021 at 14:07:06 -0500, Josef 'Jeff' Sipek wrote: Ok, just a quick update. I managed to reproduce it. I'll try to figure out where things went wro

Re: Authentication segfault with Dovecot 2.3.13

2021-01-06 Thread Aki Tuomi
It would appear that nodelay causes the crash here. We are tracking this bug as DOV-4280 internally. Aki > On 07/01/2021 01:05 Josef 'Jeff' Sipek wrote: > > > On Wed, Jan 06, 2021 at 14:07:06 -0500, Josef 'Jeff' Sipek wrote: > > Ok, just a quick update. I managed to reproduce it. I'll try

Re: Authentication segfault with Dovecot 2.3.13

2021-01-06 Thread Josef 'Jeff' Sipek
On Wed, Jan 06, 2021 at 14:07:06 -0500, Josef 'Jeff' Sipek wrote: > Ok, just a quick update. I managed to reproduce it. I'll try to figure out > where things went wrong. Two more questions: (1) Why are you using a UNION in your SQL statement? (2) Does the crash still happen if you remove the n

Re: Authentication segfault with Dovecot 2.3.13

2021-01-06 Thread Josef 'Jeff' Sipek
Ok, just a quick update. I managed to reproduce it. I'll try to figure out where things went wrong. Thanks, Jeff. On Wed, Jan 06, 2021 at 18:22:03 +0100, Harald Leithner wrote: > Hi > > Am 06.01.2021 um 18:08 schrieb Josef 'Jeff' Sipek: > > On Wed, Jan 06, 2021 at 17:13:05 +0100, Harald Leith

Re: Authentication segfault with Dovecot 2.3.13

2021-01-06 Thread Harald Leithner
Hi Am 06.01.2021 um 18:08 schrieb Josef 'Jeff' Sipek: On Wed, Jan 06, 2021 at 17:13:05 +0100, Harald Leithner wrote: Hi, and the user part with "user@redacted". I assume you did this for additional privacy, and not because you think that auth_debug_passwords=no should hide usernames as w

Re: Authentication segfault with Dovecot 2.3.13

2021-01-06 Thread Harald Leithner
Hi, small correction, the devices that tries to login and triggers the segfault uses an old password and can't login and segfaults the auth process. On 2.2 the login wasn't successful too but didn't segfault. Harald Am 06.01.2021 um 18:01 schrieb John Stoffel: "Harald" == Harald Leithner w

Re: Authentication segfault with Dovecot 2.3.13

2021-01-06 Thread Josef 'Jeff' Sipek
On Wed, Jan 06, 2021 at 17:13:05 +0100, Harald Leithner wrote: > Hi, > > we have a problem upgrading Dovecot from 2.2 to 2.3.13 on one server it > seems one user triggers a segfault while trying to authenticate. > > The user works fine with 2.2.36.4, our last update try mid-late 2020 was > abor

Re: Authentication segfault with Dovecot 2.3.13

2021-01-06 Thread John Stoffel
> "Harald" == Harald Leithner writes: Harald> we have a problem upgrading Dovecot from 2.2 to 2.3.13 on one Harald> server it seems one user triggers a segfault while trying to Harald> authenticate. Can you get the user to change their password to not have quite some special characters maybe

Authentication segfault with Dovecot 2.3.13

2021-01-06 Thread Harald Leithner
Hi, we have a problem upgrading Dovecot from 2.2 to 2.3.13 on one server it seems one user triggers a segfault while trying to authenticate. The user works fine with 2.2.36.4, our last update try mid-late 2020 was aborted because of this reason. While debugging the problem we noticed that t