Details
-
Bug
-
Resolution: Unresolved
-
Major
-
None
-
None
-
Build 0.4.14-dev-13-gb3f45a2 or older.
-
0.4.16-dev-1864-g3fe5b8b https://github.com/reactos/reactos/commit/3fe5b8b0bb4617e6b804a464c1a35ce676f83c1e
Description
Noticed this while playing with OnWorks public virtual machines, as they only have ReactOS 0.4.11 release available:
- https://www.onworks.net/runos/create-os.html?os=ReactOS-0.4.11&home=init
- https://www.winfy.net/runwine/index.php?os=ReactOS-0.4.11 (mirror)
Steps to reproduce:
- Install an old ReactOS build (e.g. 0.4.13-release)
- Once installed, download the latest nightly (0.4.16-dev-1900-g7f10759 at moment of writing)
- Modify freeldr.ini RAMDisk entry to boot the bootcd.iso:
[ReactOS_Ram]BootType=ReactOSSetupSystemPath=ramdisk(0)\Options=... /RDPATH=\bootcd.iso /RDEXPORTASCD
So, 0.4.11, 0.4.12, and 0.4.13 releases are unable to RAMDisk-boot recent nightly builds, getting a bugcheck 0x7E (SYSTEM_THREAD_EXCEPTION_NOT_HANDLED) in wdfldr.sys with status 0xC0000005 (STATUS_ACCESS_VIOLATION). This problem started with 0.4.16-dev-1864-g3fe5b8b, so it's a regression.
0.4.16-dev-1863-g4207992 - booting fine
0.4.16-dev-1864-g3fe5b8b - wdfldr.sys BSOD 0x7E / 0xC0000005
Also it's hard to debug, because the builds are not talking to serial port, related to the setup process changes in later builds.
Bisecting where it started to boot newer builds fine:
0.4.14-dev-13-gb3f45a2 - wdfldr.sys BSOD 0x7E / 0xC0000005
0.4.14-dev-17-g9ad0dd1 - GPF
0.4.14-dev-20-g2f4fb90 - GPF
0.4.14-dev-22-g2d1a6b2 - setup ok, no serial
Skipped builds are either not buildable or have GPF on early boot.
Bisecting where serial debug started working:
0.4.15-dev-1260-gd45b268 - setup ok, no serial
0.4.15-dev-1261-g4792c00 - setup ok, serial ok
It was possible to get debug log output by cherry-picking changes from 0.4.15-dev-1261-g4792c00 and closely related commits on top of 0.4.14-dev-13-gb3f45a2: 0.4.16-dev-1913-gef03e09.log![]()