I'm investigating a crash on restore of a larger test case we have, in which I
have determined that the
situation is that we have a file opened (looking at /proc/<pid>/fd) that is
deleted:
30 -> /tmp/ffiWlKyqm (deleted)
Now, this has been on 1.2.4, so I tried using the latest version, but I got
stuck somewhere else,
(and haven't had a chance to figure that out).
So I tried making a small test case that could reproduce the exact error
message I was getting. Now, I
didn't get the exact same error, but I found one case that occurs when I do a
checkpoint (on both 1.2.4
and trunk). I'll still be looking for the original problem, so when I get that,
I'll be submitting another
issue to the forum.
Error shown when dmtcpCheckpoint():
[40000] ERROR at fileconnection.cpp:881 in writeFileFromFd;
REASON='JASSERT(readBytes != -1) failed'
(strerror((*__errno_location ()))) = Bad file descriptor
Message: Read Failed
delete_test (40000): Terminating...
delete_test.c:
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <dmtcpaware.h>
int main()
{
FILE *fp;
// Problematic only when in "w" mode or "a". All + modes and "r" are fine.
fp = fopen("/tmp/ff_jdl", "w");
fprintf(stdout, "Opened ff_jdl\n");
sleep(1);
fprintf(stdout, "Deleting ff_jdl\n");
unlink("/tmp/ff_jdl");
sleep(2);
dmtcpCheckpoint();
fprintf(stdout, "I have returned\n");
sleep(2);
return 0;
}
Joshua Louie
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_mar
_______________________________________________
Dmtcp-forum mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dmtcp-forum
