Post #2,101
Online now
- Joined
- 09.05.2025
- Messages
- 6,712
- Reactions
- 1,543
- Location
- United States
Thanks, that's the pair I needed. From the identifier strings in both files:
OK file —
10375563861084C531, path 37/1/EDC17C70/57/P1084C//P1084C540 → actually P1084C531, cpf V2015.46.06 / DF050, cpv V04.163.190 PVCTM / CB154, cal build dated 2014/07.Crashing file —
10375577781084C540, path 37/1/EDC17C70/57/P1084C//P1084C540, cpf V2015.46.09 / DF053, cpv V07.163.190 PVCTM / CB158, cal build 2015/02.Everything underneath is identical: same
TC1793, same EDC17_PBL_C04_50.00.01, same ME(D)/EDC17 SB_V25.00.00/1793, same TPROT_V13.00.02/1793, same 4 MB size, and both reads have the first 2 MB blank with content starting at 0x200000. So the read layout and the ECU family are not the variable here. The only thing that changes is the calibration level, C531 → C540.That makes it a cal-revision problem, not an ECU-support problem: the 14.0 EDC17C70 handler copes with the older P1084C531 build and trips on the newer P1084C540 one. Most likely the DTC/mask/path table it locates in C540 returns a bad size or an out-of-range pointer, and the read runs off into something that takes the machine down. To be clear on what's fact and what isn't: the identifiers above are read straight from the files, the handler explanation is inference — I haven't run either file through xDecoder myself.
Two things that still matter for the next step:
- The Event Viewer Critical entry at the crash time, or the minidump from
C:\Windows\Minidump. If the faulting module is a driver rather than xDecoder, that changes the fix entirely. - Whether realtime AV or endpoint protection is live. A filter driver faulting on the file read looks exactly like this, and it would explain why the box power-cycles instead of the app just closing. Worth adding an exclusion for the xDecoder folder and your bin folder and retrying C540.
Workaround in the meantime: since C531 and C540 are the same strategy and container, the DTC table format will be the same, just relocated. Load C531, note the table signature and structure, then locate the equivalent block in C540 with a hex editor and work from there. Check your checksum handling before writing anything back.
If the minidump points at xDecoder, this needs to go to Bin as an EDC17C70 cal-level report with both files attached — only a handler update fixes it properly, and I can't make that change from here.
