Latest batch of hadcm3s tasks getting quite a few failures with

<message>
process exited with code 255 (0xff, -1)</message>
<stderr_txt>
Process creation (../../projects/climateprediction.net/hadcm3s_8.36_i686-apple-darwin) failed: Bad CPU type in executable (errno = -1)

</stderr_txt>

I have been hoping someone getting this (they are all computers crashing everything they get from CPDN) might have posted asking for help and given some clues for others but nothing so far. Wondering if anyone with Mac experience can shed light on this. - Looking at several failures for the latest batch of that type, there is no pattern I can discern with respect to CPU, OS version or BOINC version. All combinations seem to have failures of this type and successes.

Hi.

My own current Mac's so far completed every task it's received — although it's having difficulty downloading one as I write, no doubt because of the IT work in Oxford.

Two of its tasks from the current batch have previously been failed with "Bad CPU type" errors on other Macs. And looking back through its previous tasks to November last year, three others have similarly failed with that error on macOS machines. I can't see any obvious reason why those other Macs should be throwing that error, but they do have a number of interesting things in common which are different from my own case:

• All have coprocessors listed.
• All are running Darwin 19.6.0.
• Four of them are using BOINC 7.14.2, the fifth BOINC 7.12.0.
• All have tended to download four tasks at a time.
• Two of them successfully completed a couple of tasks in 2019, but have barfed everything since.

My own Mac has no BOINC-usable coprocessors, runs Darwin 18.7.0 (macOS 10.14.6 Mojave) and BOINC 7.16.14, and only receives one CPDN task in any one download, although it may have two in progress at the same time. It's also configured with an increased shared memory, a hint I picked up when I first started crunching for CPDN eleven years ago.

… none of which seems to have anything to do with a bad CPU types. :\ Mind you, the full error is "Bad CPU type in executable", so maybe the "executable" part's relevant. Another wild guess would be that it's something to do with macOS security settings.

---

NG

@Nigel You got it with the Darwin version. I went through 75 work units with Mac successes. All of those successful Macs were running Darwin 18.7 or earlier. Looking at those same work units, if a Mac in those work units failed with "Bad CPU type", that Mac was running Darwin 19 or 20.

Thanks George and Nigel,

As is often the case, once you know, the answer it is obvious!

Dave Jackson wrote:

As is often the case, once you know, the answer it is obvious!

Duh! I was so close. The need to continue using a couple of old 32-bit applications is the very reason I haven't progressed my iMac beyond macOS Mojave. It just never occurred to me back in August that the CPDN executable was 32-bit only, since Macs have been able to run 64-bit software for some years. And I'd forgotten about this thread when I posted to the "New work Discussion" topic recently. Sorry. :\

---

NG

BOINC developer Charlie Fenton is BOINC's Macintosh specialist, and I'm sure would be willing to explain recent changes in Mac software and hardware to the project management team. Unfortunately, he tends to talk in terms of Mac OS X versions, which are are given different ID numbers from the Darwin numbers reported by BOINC.

I don't think he's produced a comprehensive guide for project programmers, but here's one that caught my eye:

Don't allow 32-bit project applications if running on Mac OS 10.15 or later

(from https://github.com/BOINC/boinc/pull/3426, January 2020)

The next round of changes came with the release of the M1 range of Macintosh computers running on ARM processors ("Apple Silicon"). These need an emulation layer - 'Rosetta' - to run x64 programs. https://github.com/BOINC/boinc/pull/4406 should ensure that Rosetta is offered to BOINC users running these most recent variants.

Have a periodic look at BOINC's 'download all' page to check which version of BOINC is recommended for various OS X versions.

Richard Haselgrove wrote:

Unfortunately, he tends to talk in terms of Mac OS X versions, which are are given different ID numbers from the Darwin numbers reported by BOINC.

Just to point out that this is a peculiarity of BOINC rather than of Charlie. :) Darwin is just a part of macOS.

---

NG

Fair comment, but it does make it difficult to correlate Charlie's (excellent) advice with ordinary BOINC users' reports.

This may be too much information, but... https://en.wikipedia.org/wiki/Darwin_(operating_system)#Release_history