It could be a libcurl error, there is a new one out there!

The crash address is always within the address space of libcurl.dll so it could be that.

It's a bit of a puzzle why the problem first appeared in early November though on a large number of hosts. It wasn't an update from Microsoft. And no one has owned up to changing anything on the server-side code around that time.

I had this issue since Boinc v4.45 (unhandled exception error), now code was changed, 5.8.0/1 similar error code such as 5.4.11 /5.2.13

It could be a libcurl error, there is a new one out there!

Rom has sent the stack traces through to the libCurl people as it is their problem. :-)

So thanks so far for reporting this. Hopefully it's fixed in the next release (if they get a new libCurl in time).

---

At 5.4.11, I was crashing every night. Since upgrading to developer version 5.8.4, everything seems to be stable again. There is a new libCurl included, I believe.

I have two Win XP machines running at 3.0 and 3.8 GHz. This was happening on both of those. I have a Win XP machine running at 2.0 GHz and it was not happening on that one. It was not happening on my three OS 10.4.8 Macs. Downloading and installing BOINC 5.8.7 has fixed the problem for the past three nights. Now I see 5.8.8 is available and I'm trying that on one of them. Running Rosetta only.

I still haven't seen any libCurl crashes in BOINC since the libCurl version was updated.

But BOINC on one host keeps crashing with something else now. I'm only seeing this on one host.


Unhandled Exception Detected...

- Unhandled Exception Record -
Reason: Access Violation (0xc0000005) at address 0x009B6580 write attempt to address 0x00000000

Engaging BOINC Windows Runtime Debugger...


The host crashes at the same address every time in 5.8.8, the same address again in 5.8.4 and a slightly different address in 5.8.3.

Oddly, when BOINC does crash, it doesn't write a full debug log to stderrdae.txt. The process itself remains in the process list but it's dead and it has to be manually killed before it can be restarted.

There's nothing in stdoutdae.txt to suggest what BOINC was doing just before it crashed. The last entry is 23 mins before the crash when BOINC was trying (and failing) to get work from LHC. I've disabled LHC on that host to see if it makes any difference.

---

The same host has now crashed running 5.8.11 too, with the same symptoms (boinc.exe still resident but dead, incomplete debug log written)


Unhandled Exception Detected...

- Unhandled Exception Record -
Reason: Access Violation (0xc0000005) at address 0x009B657C write attempt to address 0x00000000

Engaging BOINC Windows Runtime Debugger...

********************

BOINC Windows Runtime Debugger Version 5.8.11

Dump Timestamp : 02/12/07 08:05:19

---