Well that wasn't good. Task on the Duron with ME finished then had an error Exit status -1 (0xffffffffffffffff), with "ERROR: Couldn't rename" at the begining
and
"Writing output ...
2008-12-03 07:38:58.9200 [CRITICAL]: Failed to rename HoughFStat file to "../../projects/einstein.phys.uwm.edu/h1_0575.05_S5R4__88_S5R4a_1_0": 2: No such file or directory"
and
"2008-12-03 07:38:59.3599 [normal]: WARNING: is_zipped() couldn't open output file '../../projects/einstein.phys.uwm.edu/h1_0575.05_S5R4__88_S5R4a_1_0' (-1)", at the end.
Unless someone else has already completed a task with ME, could be possible it may not have liked being started with the stock. Or deleting the .cpt file was a problem. I'm thinking I should run one from the begining with 6.07
Unless someone else has already completed a task with ME, could be possible it may not have liked being started with the stock. Or deleting the .cpt file was a problem. I'm thinking I should run one from the begining with 6.07.
Don't. It will probably happen to all of them. What doesn't work on 98/ME is the rename() function, regardless of being used for the checkpoint or the final output file.
Thanks for testing, I'd suggest you switch back to the default 6.04 on those boxes before wasting more time on a task.
If you don't want to lose how much you crunched so far, stop the BOINC Client, manually rename the .cpt.tmp file to .cpt, manually download the 6.04*_0 and 6.04*_1 files, rename them to 6.07*, make a copy of 6.07*_1 as 6.07_2, replace the 6.07 files in the project folder with the renamed 6.04 files (the graphics app is the same anyway). Then start the Client again.
I need some confirmation from the people that had problems with checkpoint renaming with the 6.05 App that their problem is solved now with 6.07. Once I could report this to BOINC, we'll fix this in the BOINC code and then I'll build a new App version from that.
1) Thanks for the warning - I'll switch the Windows 98 box back to 6.04 before it tries to rename the output file (unless you're likely to have a fix in place before it finishes, in 8 days' time?)
2) I have installed 6.07 on an SSE2 XP host, and it's renaming correctly - but it wasn't one of the ones that had problems in the first place.
This is mainly a bugfix release (candidate) for the rename() problem on Win98/ME etc. It is a "maximum compatibility" App not primarily designed for speed to make sure it runs on old systems running Win98.
Please try this on the Win98 systems that had problems with 6.07 and also make sure it works on newer systems, too (i.e. without "can't rename" warnings etc.). If this test is successful, I'll build a new full-featured App from that code.
v6.08 is installed on the Win 98 SE machine, and seems to checkpoint OK - I haven't been watching closely enough to see exactly how it does it, but the .CPT file is there, with a regularly changing timestamp, and no .TMP files in sight.
I have a slight suspicion that stderr.txt is not being kept fully up-to-date, and I have one small but beneficial side-effect to report: I monitor the hosts remotely using BoincView, and recent Einstein builds haven't reported progress %age uniformly to BV: the progress counter jumps up a full quantum at the checkpoint intervals. This app, while not reporting absolutely smooth progress, updates the %age done more frequently than once per checkpoint. [It makes a difference on this box - the checkpoints are almost 15 minutes apart! I had a minor panic when the %age moved, but no checkpoint file appeared].
Similar experience on my ME machine. Seems to be running well, and the progress counter does update more often as well (thought I noticed that on the 6.07). I had just finished a 6.04 and was only a little over 1% into this task, so this will be almost full 6.08
The more frequent progress update is a feature that we added in preparation for S5R5 where the time the App will look at one sky location will be much longer than in S5R4 (and before). Previously the progress counter was only updated with every checkpoint, which is not the case anymore.
and also make sure it works on newer systems, too (i.e. without "can't rename" warnings etc.).
Just finished this 6.08 WU on the Intel E2180 dual core as requested. Seems to have run smoothly with no problems to report. About 19 and 3/4 hours crunch time.
Quote:
It is a "maximum compatibility" App not primarily designed for speed to make sure it runs on old systems running Win98.
You weren't kidding lol, my ME maching is almost 36 hours in and has about 4 hours more to go. Makes the 31 hours it was taking on stock attractive ;P
Well that wasn't good. Task
)
Well that wasn't good. Task on the Duron with ME finished then had an error Exit status -1 (0xffffffffffffffff), with "ERROR: Couldn't rename" at the begining
and
"Writing output ...
2008-12-03 07:38:58.9200 [CRITICAL]: Failed to rename HoughFStat file to "../../projects/einstein.phys.uwm.edu/h1_0575.05_S5R4__88_S5R4a_1_0": 2: No such file or directory"
and
"2008-12-03 07:38:59.3599 [normal]: WARNING: is_zipped() couldn't open output file '../../projects/einstein.phys.uwm.edu/h1_0575.05_S5R4__88_S5R4a_1_0' (-1)", at the end.
Unless someone else has already completed a task with ME, could be possible it may not have liked being started with the stock. Or deleting the .cpt file was a problem. I'm thinking I should run one from the begining with 6.07
RE: Unless someone else has
)
Don't. It will probably happen to all of them. What doesn't work on 98/ME is the rename() function, regardless of being used for the checkpoint or the final output file.
Thanks for testing, I'd suggest you switch back to the default 6.04 on those boxes before wasting more time on a task.
If you don't want to lose how much you crunched so far, stop the BOINC Client, manually rename the .cpt.tmp file to .cpt, manually download the 6.04*_0 and 6.04*_1 files, rename them to 6.07*, make a copy of 6.07*_1 as 6.07_2, replace the 6.07 files in the project folder with the renamed 6.04 files (the graphics app is the same anyway). Then start the Client again.
I need some confirmation from the people that had problems with checkpoint renaming with the 6.05 App that their problem is solved now with 6.07. Once I could report this to BOINC, we'll fix this in the BOINC code and then I'll build a new App version from that.
BM
BM
1) Thanks for the warning -
)
1) Thanks for the warning - I'll switch the Windows 98 box back to 6.04 before it tries to rename the output file (unless you're likely to have a fix in place before it finishes, in 8 days' time?)
2) I have installed 6.07 on an SSE2 XP host, and it's renaming correctly - but it wasn't one of the ones that had problems in the first place.
A new App 6.08 for Windows
)
A new App 6.08 for Windows can be found on the Power User's Apps page.
This is mainly a bugfix release (candidate) for the rename() problem on Win98/ME etc. It is a "maximum compatibility" App not primarily designed for speed to make sure it runs on old systems running Win98.
Please try this on the Win98 systems that had problems with 6.07 and also make sure it works on newer systems, too (i.e. without "can't rename" warnings etc.). If this test is successful, I'll build a new full-featured App from that code.
Thanks for your help.
BM
BM
v6.08 is installed on the Win
)
v6.08 is installed on the Win 98 SE machine, and seems to checkpoint OK - I haven't been watching closely enough to see exactly how it does it, but the .CPT file is there, with a regularly changing timestamp, and no .TMP files in sight.
I have a slight suspicion that stderr.txt is not being kept fully up-to-date, and I have one small but beneficial side-effect to report: I monitor the hosts remotely using BoincView, and recent Einstein builds haven't reported progress %age uniformly to BV: the progress counter jumps up a full quantum at the checkpoint intervals. This app, while not reporting absolutely smooth progress, updates the %age done more frequently than once per checkpoint. [It makes a difference on this box - the checkpoints are almost 15 minutes apart! I had a minor panic when the %age moved, but no checkpoint file appeared].
Similar experience on my ME
)
Similar experience on my ME machine. Seems to be running well, and the progress counter does update more often as well (thought I noticed that on the 6.07). I had just finished a 6.04 and was only a little over 1% into this task, so this will be almost full 6.08
Thanks for testing. The
)
Thanks for testing.
The more frequent progress update is a feature that we added in preparation for S5R5 where the time the App will look at one sky location will be much longer than in S5R4 (and before). Previously the progress counter was only updated with every checkpoint, which is not the case anymore.
BM
BM
RE: and also make sure it
)
Just finished this 6.08 WU on the Intel E2180 dual core as requested. Seems to have run smoothly with no problems to report. About 19 and 3/4 hours crunch time.
You weren't kidding lol, my ME maching is almost 36 hours in and has about 4 hours more to go. Makes the 31 hours it was taking on stock attractive ;P
RE: You weren't kidding
)
If it's the same download package I used it's an FPU-only test build - no _1 SSE or _2 SSE2 apps - proof-of-concept only.
Unless there's a tribe out there running Windows 98 on Nehalems?
RE: RE: You weren't
)
Yes, it is the same one. Running these test builds helps remind me how much I appreciate the work the developers put into the apps.