I seem to have gotten past the FORTRAN library problem, but now I\'m seeing this when trying to run the sulphur cycle model:

Insufficient Memory/Stack Space Available!
2005-09-26 16:58:53 [climateprediction.net] Unrecoverable error for result 47cd_b00296413_0 (process got signal 10)
2005-09-26 16:58:53 [climateprediction.net] Unrecoverable error for result 47cd_b00296413_0 (process got signal 10)

The crashing thread\'s backtrace was:
Thread 1 Crashed:
0 libGL.dylib 0x92f3f6c0 glDeleteTextures + 48
1 ...r_4.21_powerpc-apple-darwin 0x0001e23c graphics_thread_cleanup + 432 (crt.c:300)
2 ...r_4.21_powerpc-apple-darwin 0x00005a64 app_cleanup() + 36 (crt.c:300)
3 ...r_4.21_powerpc-apple-darwin 0x00006f58 checkBOINCStatus(bool) + 228 (crt.c:300)
4 ...r_4.21_powerpc-apple-darwin 0x000054dc mainLoop() + 76 (crt.c:300)
5 ...r_4.21_powerpc-apple-darwin 0x00004fbc worker() + 1844 (crt.c:300)
6 ...r_4.21_powerpc-apple-darwin 0x0004c810 foobar(void*) + 60 (graphics_impl.C:75)
7 libSystem.B.dylib 0x9002b130 _pthread_body + 108

---

The stacktrace indeicates the problem might have been in the graphics program.
Is this a repetitive behaviour?

---

The stacktrace indeicates the problem might have been in the graphics program.
Is this a repetitive behaviour?

Yes, it\'s happening every time.

---

I\'m still seeing this crash 100% on my dual G5 (the only machine it\'s happening on). The machine has been unusable for climateprediction.net for over a month due to this issue. None of my single-processor G4s are seeing this crash.

Given that it\'s happening in a thread other than the main thread, and that it\'s only happening on a dual processor machine, I think it\'s probably a concurrency error. It might be due to the weak memory synchronization model on PowerPC.

Prior versions of the climateprediction.net application ran great on this machine, so I think it was a change introduced in 4.12 or later. Wasn\'t graphics support first introduced in that version?

Is there an older version of the app that I can run using app_info.xml?

---

Sorry, you can\'t go backwards with app versions, only with BOINC.

Perhaps if you severly limit available hd space so that you get a \"not enough disk space\" message, and then gradually increase it, you can get a slab model, which is not quite as demanding as sulphur.

This crash is happening in the regular model, too, so I don\'t think that workaround would work.

Is the source for that file (crt.c) online anywhere? I\'d like to look at it and see if I can spot a concurrency problem.

---

If crt.c stands for \'chinese remainder theorem\', there is a page <a href=\"http://swtch.com/usr/local/plan9/9pm/src/libmp/crt.c\"> here.</a> :)

Google shows over 27,000 pages for crt.c, so it depends on what it\'s supposed to be part of.

But if crt.c is part of the cpdn app, then no, it\'s not available.
The software is owned by the UK Met Office, and Oxford Uni only has a license to port and develop it for pc\'s, not to make it available to the public.

If Carl and Tolu weren\'t so busy with \'spinup\', they might look at it for you.

P.S. Have you read <a href=\"http://www.climateprediction.net/newsb.php?id=1\"> this</a> about needing extra files?

Well, clearly, I meant the crt.c that is in the crash backtrace, which also, clearly, is part of the cpdn application. Since the file in question seems to be BOINC-specific (what with function names like checkBOINCStatus), and not part of the actual scientific code, I thought it might be available. Perhaps not. If it is available, I thought I might be able to track down the concurrency issue that seems to be causing the crash. PowerPC has a weak memory model that can cause crashes where the same code works on x86 multiprocessor machines.

Yes, I have installed the necessary libraries. That\'s not the issue here.

---

Why is a concurrency error happening with CPDN, but not other BOINC projects? Does it have to do with the particulars of this project, or the availability/knowledge of Mac programmers on the CPDN team?

Would it process any more successfully those of us with dual processor machines changed our account settings to say our machine has 1 processor?

---

G5 dual 2GHz OSX 10.3.9
G4 dual 867MgHz OSX 10.4.3

The crash is happening in the CPDN application(s), so it\'s a particular bug in that application. That\'s why it\'s not happening in other BOINC apps. It\'s no reflection on the CPDN programmers; the PowerPC has a memory architecture that\'s harder to deal with than some other platforms. It\'s tripped up lots of good programmers.

If it really is a concurrency problem (and that\'s speculation on my part), setting one processor in your BONC prefs will not help. That just affects how many BOINC apps you run simultaneously. The CPDN app is using multiple threads, and even if you say \"just use one processor\" those threads can still run on different processors, which is what can cause such a problem. To stop the problem, you\'d have to actually disable one of your processors (which you can do from the command line by changing NVRAM or using the Developer Tools).

There was another post by someone who says they saw this on a single-processor G5 (iMac G5), so it\'s possible it\'s not a concurrency error.

---