[MDEV-22555] 10.4.13 MSI install or upgrade fails Created: 2020-05-14 Updated: 2020-08-11 Resolved: 2020-05-15 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Server |
| Affects Version/s: | 10.4.13 |
| Fix Version/s: | 10.4.14 |
| Type: | Bug | Priority: | Critical |
| Reporter: | Toshimitsu Nakanishi | Assignee: | Vladislav Vaintroub |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | Windows, not-10.5, packaging | ||
| Environment: |
Windows 10(x64) |
||
| Attachments: |
|
||||||||||||||||
| Issue Links: |
|
||||||||||||||||
| Description |
|
When completing install mariadb-10.4.13-winx64.msi. 10.4.12 was OK. Installed OS was clean. It does not have Visual Studio CPP runtime. Please, check it. |
| Comments |
| Comment by Vladislav Vaintroub [ 2020-05-14 ] | ||||||
|
viewhero, can you send the installation log, please? Did you upgrade on top of 10.4.12? Did installation of the latest VS2019 runtime help? | ||||||
| Comment by Toshimitsu Nakanishi [ 2020-05-14 ] | ||||||
|
Hi, Vladislav. I tested 3 environments. 1. upgrade from 10.4.12. PC has Visual Studio 2017&2019 . 2. upgrade from 10.4.12. None of Visual Studio. 3. new install 10.4.13. Clearn PC. Please, check it. | ||||||
| Comment by Vladislav Vaintroub [ 2020-05-14 ] | ||||||
|
well, the reboot prompt in finish.jpg, or in the MSI log, it is MSIs own magic, that we can't prevent. It is not a bug by itself, you can choose to continue and restart. You stopped the installation before it became "interesting" Failure for the mariadb service to start, that's bug. I can already guess where it is, we used some outdated merge modules that did not include | ||||||
| Comment by Vladislav Vaintroub [ 2020-05-14 ] | ||||||
|
https://github.com/MariaDB/server/commit/3bfe305c5cd678a8563f7a76d6ed59095129007e will probably solve that | ||||||
| Comment by Vladislav Vaintroub [ 2020-05-14 ] | ||||||
|
viewhero , I put a new build of mariadb MSI to my OneDrive share , here https://1drv.ms/u/s!AKT4tBcElQz_j5Ej I'd appreciate if you can test it on clean PC, as I do not have anything like that clean PC right at the moment. | ||||||
| Comment by Toshimitsu Nakanishi [ 2020-05-15 ] | ||||||
|
Thank you for uploading fixed version. I test 2 environment. Both of tests are OK. This problem is resolved. | ||||||
| Comment by Vladislav Vaintroub [ 2020-05-26 ] | ||||||
|
For everyone in the future who googles "MariaDB 10.4.13 MSI fails" here is The Resolution
| ||||||
| Comment by Vladislav Vaintroub [ 2020-05-26 ] | ||||||
|
changed title to be better googleable | ||||||
| Comment by Ilguiz Latypov [ 2020-08-09 ] | ||||||
|
The installer refuses to run on Windows Server 2008, saying "The application is only supported on Windows 10 and Windows 2016". Downloading and unpacking the ZIP shows the error 0xc0000022 (STATUS_ACCESS_DENIED).
mariadb-admin.exe and other tools work. | ||||||
| Comment by Ilguiz Latypov [ 2020-08-09 ] | ||||||
|
As I used Cygwin's unzip, it assigned mis-ordered Windows ACL permissions to the files. But fixing the permissions, including the recursive inheritance, giving myself full control, did not resolve the issue. Attaching the process monitor with a filter "process name includes mariadbd". The full log shows some BUFFER_OVERFLOW result in Symantec antivirus trying to read a registry value around the same time.
| ||||||
| Comment by Ilguiz Latypov [ 2020-08-09 ] | ||||||
|
Perhaps, it's about Symantec's failing to validate Microsoft's SHA-2 signatures. https://www.zdnet.com/article/symantec-cannot-handle-sha-2-and-breaks-windows-7-and-server-2008-r2/ | ||||||
| Comment by Ilguiz Latypov [ 2020-08-09 ] | ||||||
|
My naive stripping the files in bin and lib of their signatures using https://github.com/ng-pe/RemoveSignCode possibly got me further where I got another error
The event log had this in the System folder,
| ||||||
| Comment by Vladislav Vaintroub [ 2020-08-09 ] | ||||||
|
See, there is an actual reason why 10.5 msi did not want to install on EOLed Windows. I also think you took unrelated bug report for commenting eelgheez | ||||||
| Comment by Ilguiz Latypov [ 2020-08-09 ] | ||||||
|
Sorry to hijack the solved issue. (But the not-in-my-backyard vision of the older Windows versions made me step away from trying MariaDB, sorry. I doubt that getting a system time could not be interpolated for the older Windows). | ||||||
| Comment by Vladislav Vaintroub [ 2020-08-10 ] | ||||||
|
It is not about getting system time. It is about getting precise system time. We would not use that function GetSystemTimePreciseAsFileTime just for the sake of forcing you off the obsolete Windows version (abandoned by its vendor), because that function is also more expensive than the non-precise variation GetSystemTimeAsFileTime we used previously. |