If I read correctly, it is only output for the compile step which is output to compilelog.txt and thus should be closed by the shell by the time xenofarm.sh is done. Oddly enough things started to work with a new tar. Might be buggy tar or buggy OS. The latter more probable I suppose.
Now there are results from that machine on pikefarm, but the testsuite fails in two cases (the same) on 7.6 and 7.7, one is some library issue with GL, test_resolve.pike segfaults. The other is in modules/files/ and the socktest.pike test.
For 7.4 it fails with the same two C stack overflows as some other architechures.
Mikael
Henrik GrubbstrXm (Lysator) @ Pike (-) developers forum wrote:
That should not be the case here, since compilelog.txt is created by the xenofarm.sh script, which should have terminated by the time tar is called.
If I remember correctly, compilelog.txt is stdout for sections of the script. It could be that some backgrounded script still holds it open.
/ Henrik Grubbström (Lysator)
Previous text:
2004-08-23 11:11: Subject: Re: Pikefarm client problem
That should not be the case here, since compilelog.txt is created by the xenofarm.sh script, which should have terminated by the time tar is called.
// Mikael
Fredrik (Naranek) Hubinette (T minus 2 weeks) @ Pike (-) developers forum wrote:
Maybe because stdout/stderr from tar is sent to the file it is trying to tar up?
/ Fredrik (Naranek) Hubinette (T minus 2 weeks)
Previous text:
2004-08-22 17:10: Subject: Pikefarm client problem
Hm, why would the file change while being read? It's not running two xenofarms at the same time or something (it shouldn't happen anyway, iirc, but...).
Or is tar just buggy?
/ Mirar
/ Brevbäraren