Open e14f4152-4982-4ace-8c95-73a0599b109b opened 14 years ago
Is this error reproducible, does it happen every time? What happens if you run sage -gp
on that machine?
Reviewer: Jeroen Demeyer
Nobody seems to care very much, so I'm closing this as invalid.
I have had this result in a doctest failure onOpenSolaris on more than one occasion. But it's not every reproducible. The test fails, then passes when rerun.
Sorry I did not comment earlier.
Dave
One "solution" to track down further this problem is to change the exception handling code in sage/interfaces/expect.py
. This is the relevant code:
try:
if self.__remote_cleaner and self._server:
c = 'sage-native-execute ssh %s "nohup sage -cleaner" &'%self._server
os.system(c)
self._expect = pexpect.spawn(cmd, logfile=self.__logfile)
if self._do_cleaner():
cleaner.cleaner(self._expect.pid, cmd)
except (ExceptionPexpect, pexpect.EOF, IndexError):
self._expect = None
self._session_number = BAD_SESSION
failed_to_start.append(self.__name)
raise RuntimeError, "Unable to start %s because the command '%s' failed.\n%s"%(
self.__name, cmd, self._install_hints())
The problem is that we probably lose information by catching the exception and re-raising a different exception. We only know that gp
failed, but not for which reason. I'm guessing something OS-related like not enough memory, but we don't know for sure.
Replying to @jdemeyer:
One "solution" to track down further this problem is to change the exception handling code in
sage/interfaces/expect.py
. This is the relevant code:try: if self.__remote_cleaner and self._server: c = 'sage-native-execute ssh %s "nohup sage -cleaner" &'%self._server os.system(c) self._expect = pexpect.spawn(cmd, logfile=self.__logfile) if self._do_cleaner(): cleaner.cleaner(self._expect.pid, cmd) except (ExceptionPexpect, pexpect.EOF, IndexError): self._expect = None self._session_number = BAD_SESSION failed_to_start.append(self.__name) raise RuntimeError, "Unable to start %s because the command '%s' failed.\n%s"%( self.__name, cmd, self._install_hints())
The problem is that we probably lose information by catching the exception and re-raising a different exception. We only know that
gp
failed, but not for which reason. I'm guessing something OS-related like not enough memory, but we don't know for sure.
I doubt it's a lack of memory. My machine has 12 GB RAM and I suspect 50-100 GB or so of swap. That said, others do use the machine, but I usually do check logs when I get doctest failures - that's why I wish the date+time of the doctests were recorded.
I know it's not normal to reopen tickets, but since this one was closed only a few hours ago, with no releases of Sage since, would it be sensible to reopen this? The problem does still exist and on more than one platform
Dave
Changed reviewer from Jeroen Demeyer to none
Minh Nguyen gets this error on a GCC compile farm Debian 5.0 machine (gcc100.fsffrance.org, AMD Opteron(tm) Processor 252 @ 2647.708 MHz) with a trial "final" 4.5.3 (essentially the same as 4.5.3.rc0):
Here is the test log.
Here is the only mention I found of a similar message on sage-devel.
CC: @sagetrac-mvngu
Component: interfaces
Issue created by migration from https://trac.sagemath.org/ticket/9866