hi,
looking at the results of pikefarm for my indy, it appears that the build process tries to do to many things at once and my machine can't cope with it.
is there any way the build process can be made less taxing on the machine? there is hardly anything else running on it.
greetings, martin.
looking at the results of pikefarm for my indy, it appears that the build process tries to do to many things at once and my machine can't cope with it.
As far as I can see, it looks like it runs out of memory during configure; it didn't do this with the builds from a few weeks ago.
How much memory and swap does the machine have? Are there any hanged processes on the machine? Is /tmp or /usr/tmp ram filesystems? If they are, is there a lot of data in them? Does it fail in the same manner if you try to build by hand? If it does, can you try adding a set -x to the main configure-script after the block of echos starting with "Compiler summary:"?
is there any way the build process can be made less taxing on the machine? there is hardly anything else running on it.
configure is about as light weight as it gets...
/ Henrik Grubbström (Lysator)
Previous text:
2004-03-18 20:54: Subject: pikefarm problems on irix 6.2
hi,
looking at the results of pikefarm for my indy, it appears that the build process tries to do to many things at once and my machine can't cope with it.
is there any way the build process can be made less taxing on the machine? there is hardly anything else running on it.
greetings, martin.
/ Brevbäraren
On Fri, Mar 19, 2004 at 07:25:01PM +0100, Henrik Grubbström (Lysator) @ Pike (-) developers forum wrote:
How much memory and swap does the machine have?
48MB ram $ swap -l lswap path dev pri swaplo blocks free maxswap vswap 2 /disk5/.swap.virtual 128,87 3 0 409600 7728 409600 0 1 /dev/swap 128,17 0 0 81896 0 81896 0
ok, swap is full. i know i had to reboot the machine every 2 weeks while it was in active use at the university, now it has been running for 60 days. short of any other sugesstions to clear swap i'll reboot now.
Are there any hanged processes on the machine?
no.
Is /tmp or /usr/tmp ram filesystems?
no.
Does it fail in the same manner if you try to build by hand?
didn't try yet (am at the CeBit now, will try next week)
greetings, martin.
48MB ram
Should be ok if you also have sufficient swap.
$ swap -l lswap path dev pri swaplo blocks free maxswap vswap 2 /disk5/.swap.virtual 128,87 3 0 409600 7728 409600 0 1 /dev/swap 128,17 0 0 81896 0 81896 0
Ok, so ~240MB swap?
ok, swap is full. i know i had to reboot the machine every 2 weeks while it was in active use at the university, now it has been running for 60 days. short of any other sugesstions to clear swap i'll reboot now.
Have you checked if there's any process that has eaten all the memory?
Does it fail in the same manner if you try to build by hand?
didn't try yet (am at the CeBit now, will try next week)
Since you were out of swap, building by hand would most likely fail in the same way.
greetings, martin.
/ Henrik Grubbström (Lysator)
Previous text:
2004-03-20 15:37: Subject: Re: pikefarm problems on irix 6.2
On Fri, Mar 19, 2004 at 07:25:01PM +0100, Henrik Grubbström (Lysator) @ Pike (-) developers forum wrote:
How much memory and swap does the machine have?
48MB ram $ swap -l lswap path dev pri swaplo blocks free maxswap vswap 2 /disk5/.swap.virtual 128,87 3 0 409600 7728 409600 0 1 /dev/swap 128,17 0 0 81896 0 81896 0
ok, swap is full. i know i had to reboot the machine every 2 weeks while it was in active use at the university, now it has been running for 60 days. short of any other sugesstions to clear swap i'll reboot now.
Are there any hanged processes on the machine?
no.
Is /tmp or /usr/tmp ram filesystems?
no.
Does it fail in the same manner if you try to build by hand?
didn't try yet (am at the CeBit now, will try next week)
greetings, martin.
/ Brevbäraren
On Sat, Mar 20, 2004 at 04:10:03PM +0100, Henrik Grubbström (Lysator) @ Pike (-) developers forum wrote:
lswap path dev pri swaplo blocks free maxswap vswap 128,87 3 0 409600 7728 409600 0 128,17 0 0 81896 0 81896 0
Ok, so ~240MB swap?
yes. (i just learned about the swap command today :-) $ swap -ln # path pri pswap free maxswap vswap 2 /disk5/.swap.virtual 3 200.00m 200.00m 200.00m 0.00k 1 /dev/swap 0 39.99m 39.99m 39.99m 0.00k
Have you checked if there's any process that has eaten all the memory?
not explicitly, but other than xenofarm nothing is running on the machine.
Since you were out of swap, building by hand would most likely fail in the same way.
i suppose so. i rebooted now :-(
is there a way to force xenofarm to rebuild? could xenofarm be adapted to recognize transient errors which are unrelated to source changes?
greetings, martin.
is there a way to force xenofarm to rebuild?
Yes, just remove pike*/<hostname>/last_*. That will not give your new result parsed by the result-parser server-side though. It only parses one result per build and node.
could xenofarm be adapted to recognize transient errors which are unrelated to source changes?
The parser can always be extended with more heuristics, but in most cases it is probably more effient to let a real person look at it.
/ Peter Bortas (Kein Paket!)
Previous text:
2004-03-20 16:56: Subject: Re: pikefarm problems on irix 6.2
On Sat, Mar 20, 2004 at 04:10:03PM +0100, Henrik Grubbström (Lysator) @ Pike (-) developers forum wrote:
lswap path dev pri swaplo blocks free maxswap vswap 128,87 3 0 409600 7728 409600 0 128,17 0 0 81896 0 81896 0
Ok, so ~240MB swap?
yes. (i just learned about the swap command today :-) $ swap -ln # path pri pswap free maxswap vswap 2 /disk5/.swap.virtual 3 200.00m 200.00m 200.00m 0.00k 1 /dev/swap 0 39.99m 39.99m 39.99m 0.00k
Have you checked if there's any process that has eaten all the memory?
not explicitly, but other than xenofarm nothing is running on the machine.
Since you were out of swap, building by hand would most likely fail in the same way.
i suppose so. i rebooted now :-(
is there a way to force xenofarm to rebuild? could xenofarm be adapted to recognize transient errors which are unrelated to source changes?
greetings, martin.
/ Brevbäraren
It would be nice if new results would show up for the same build.
/ Mirar
Previous text:
2004-03-20 17:06: Subject: Re: pikefarm problems on irix 6.2
is there a way to force xenofarm to rebuild?
Yes, just remove pike*/<hostname>/last_*. That will not give your new result parsed by the result-parser server-side though. It only parses one result per build and node.
could xenofarm be adapted to recognize transient errors which are unrelated to source changes?
The parser can always be extended with more heuristics, but in most cases it is probably more effient to let a real person look at it.
/ Peter Bortas (Kein Paket!)
Yes. Not so nice that _I_ will do the extra coding though. I'f one does that it's time to start looking at the security aspect too. Clients should have certificates.
/ Peter Bortas (Kein Paket!)
Previous text:
2004-03-20 17:13: Subject: Re: pikefarm problems on irix 6.2
It would be nice if new results would show up for the same build.
/ Mirar
Is there a developer setup of xenofarm? I don't think I would mind coding on it now and then, but...
/ Mirar
Previous text:
2004-03-20 17:15: Subject: Re: pikefarm problems on irix 6.2
Yes. Not so nice that _I_ will do the extra coding though. I'f one does that it's time to start looking at the security aspect too. Clients should have certificates.
/ Peter Bortas (Kein Paket!)
No, but it's pretty easy to set up. Check out xenofarm, add the Roxen module to a Roxen 2.4+, point the module to a mysql it can use. Then run the package creation script when you want new packages, and run the parser script when you have results waiting in the inbox.
/ Peter Bortas (Kein Paket!)
Previous text:
2004-03-20 17:16: Subject: Re: pikefarm problems on irix 6.2
Is there a developer setup of xenofarm? I don't think I would mind coding on it now and then, but...
/ Mirar
pike-devel@lists.lysator.liu.se
-
Henrik Grubbstr�m (Lysator) ＠ Pike (-) developers forum -
Martin Baehr -
Martin Bähr
-
Mirar ＠ Pike developers forum
-
Peter Bortas (Kein Paket!) ＠ Pike (-) developers forum