Re: Defunct processes with 1.5.25-15; seemingly reproducible
On Sat, Nov 28, 2009 at 01:04:14PM -0500, Dave Steenburgh wrote: >I "fixed" the problem. I have not updated any software nor changed >any scripts, and yet I can no longer reproduce these behaviors. I am >absolutely certain that there is more than one problem here, and I >hope someone takes a close look at my previous posts. Once again: defunct processes are not necessarily a symptom of a Cygwin bug. However, even if they are, the 1.5 series of Cygwin is now a dead-end. There won't be any future fixes there. cgf -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Re: Defunct processes with 1.5.25-15; seemingly reproducible
I "fixed" the problem. I have not updated any software nor changed any scripts, and yet I can no longer reproduce these behaviors. I am absolutely certain that there is more than one problem here, and I hope someone takes a close look at my previous posts. -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Re: Defunct processes with 1.5.25-15; seemingly reproducible
Dave Steenburgh wrote: > And what is the significance of the duplicated lines that ps showed > me? Each of three processes did indeed have a lock on an output file > from the most recent session, and yet all the information that ps > showed was identical to a previous session. I started a new session just minutes ago. The process that parses my data and pipes commands to gnuplot happened to have a PID of 5240. ps indicated that PID 5240 was defunct, and all the other information appears the same (though the timestamp is now a date). Although the process was not actually defunct, I killed that session and started a new one. The process with the same name of course has a new PID, but it is /not/ listed as defunct. Whatever is the cause of the first defunct processes, it seems [to me] that the more recent ones have a different cause. -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Re: Defunct processes with 1.5.25-15; seemingly reproducible
Paul McFerrin wrote: > A process is not necesarily the real "problem". All > process are processes that their parent has NOT done a wait(2) yet. Since > these processes have called exit(2), they must hang around until a > wait(2) is completed so that the exit status can be returned to the parent. > > You need to understand more of the parent/child relationship before > you can pass blame, if any. Personally, I'm not trying to blame anyone. I'm trying to figure out what the real problem is, but I'm not familiar enough with the details to track down the real problem. How about this...is there any way that I can be absolutely certain which process was the parent? And what is the significance of the duplicated lines that ps showed me? Each of three processes did indeed have a lock on an output file from the most recent session, and yet all the information that ps showed was identical to a previous session. -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Re: Defunct processes with 1.5.25-15; seemingly reproducible
A process is not necesarily the real "problem". All process are processes that their parent has NOT done a wait(2) yet. Since these processes have called exit(2), they must hang around until a wait(2) is completed so that the exit status can be returned to the parent. You need to understand more of the parent/child relationship before you can pass blame, if any. Dave Steenburgh wrote: cgf wrote: Defunct processes are not necessarily indicative of a cygwin problem. This could easily be a problem with gnuplot. Given the sum of my own limited knowledge of the problem at hand (in summary: every program involved is in my local cygwin directory), I figured it was best to ask here first. Dave Steenburgh wrote: $ ps PIDPPIDPGID WINPID TTY UIDSTIME COMMAND ... 416412887684 55046 1003 23:49:20 539232245984 61005 1003 23:49:06 145252405984 81045 1003 23:49:06 524032245984 45325 1003 23:49:06 ... ... The PIDs seem to be the same every time this happens. Specifically, I have seen 5240 and 1452 every time. Whether that's significant, I don't know. Now, this is interesting... I killed those defunct processes with process explorer, and subsequently confirmed via ps, task manager, and process explorer that they were no longer running. I began another session with gnuplot, and this time there are three: $ ps ... 524032245984 45325 1003 23:49:06 416412887684 55046 1003 23:49:20 539232245984 61005 1003 23:49:06 ... For each of those, the entire row is identical to a row in my previous message. Since I started the session around 14:00, and the last output file's modification timestamp is 17:21, I'd say the timestamps for those three processes are not reliable. Is it possible that the original defunct processes were never truly killed? If so, can they be killed without rebooting? -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Re: Defunct processes with 1.5.25-15; seemingly reproducible
2009/11/27 Dave Steenburgh > > cgf wrote: > > Defunct processes are not necessarily indicative of a cygwin problem. > > This could easily be a problem with gnuplot. > > Given the sum of my own limited knowledge of the problem at hand (in > summary: every program involved is in my local cygwin directory), I > figured it was best to ask here first. > > Dave Steenburgh wrote: > > $ ps > > PID PPID PGID WINPID TTY UID STIME COMMAND > > ... > > 4164 1288 7684 5504 6 1003 23:49:20 > > 5392 3224 5984 6100 5 1003 23:49:06 > > 1452 5240 5984 8104 5 1003 23:49:06 > > 5240 3224 5984 4532 5 1003 23:49:06 > > ... > > > ... > > The PIDs seem to be the same every time this happens. Specifically, I > > have seen 5240 and 1452 every time. Whether that's significant, I > > don't know. > > Now, this is interesting... I killed those defunct processes with > process explorer, and subsequently confirmed via ps, task manager, and > process explorer that they were no longer running. I began another > session with gnuplot, and this time there are three: > > $ ps > ... > 5240 3224 5984 4532 5 1003 23:49:06 > 4164 1288 7684 5504 6 1003 23:49:20 > 5392 3224 5984 6100 5 1003 23:49:06 > ... > > For each of those, the entire row is identical to a row in my previous > message. Since I started the session around 14:00, and the last > output file's modification timestamp is 17:21, I'd say the timestamps > for those three processes are not reliable. Is it possible that the > original defunct processes were never truly killed? If so, can they > be killed without rebooting? > Those defunct function will disapear after parent process exist in *nix. To avoid this condition, try to handle SIGCHLD and call wait/waitpid to free resources used by child process -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Re: Defunct processes with 1.5.25-15; seemingly reproducible
cgf wrote: > Defunct processes are not necessarily indicative of a cygwin problem. > This could easily be a problem with gnuplot. Given the sum of my own limited knowledge of the problem at hand (in summary: every program involved is in my local cygwin directory), I figured it was best to ask here first. Dave Steenburgh wrote: > $ ps > PIDPPIDPGID WINPID TTY UIDSTIME COMMAND > ... > 416412887684 55046 1003 23:49:20 > 539232245984 61005 1003 23:49:06 > 145252405984 81045 1003 23:49:06 > 524032245984 45325 1003 23:49:06 > ... > ... > The PIDs seem to be the same every time this happens. Specifically, I > have seen 5240 and 1452 every time. Whether that's significant, I > don't know. Now, this is interesting... I killed those defunct processes with process explorer, and subsequently confirmed via ps, task manager, and process explorer that they were no longer running. I began another session with gnuplot, and this time there are three: $ ps ... 524032245984 45325 1003 23:49:06 416412887684 55046 1003 23:49:20 539232245984 61005 1003 23:49:06 ... For each of those, the entire row is identical to a row in my previous message. Since I started the session around 14:00, and the last output file's modification timestamp is 17:21, I'd say the timestamps for those three processes are not reliable. Is it possible that the original defunct processes were never truly killed? If so, can they be killed without rebooting? -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Re: Defunct processes with 1.5.25-15; seemingly reproducible
On Thu, Nov 26, 2009 at 01:52:45PM -0500, Dave Steenburgh wrote: >It is my understanding that this problem is not easily reproducible. >Well, I've been reproducing it locally since last night. ?I'm going to >try leaving every cygwin-related process as-is as long as necessary, >in the hope of beating this problem into submission. > >The problem: > >$ ps >? ? ?PID ? ?PPID ? ?PGID ? ? WINPID ?TTY ?UID ? ?STIME COMMAND >... >? ? 4164 ? ?1288 ? ?7684 ? ? ? 5504 ? ?6 1003 23:49:20 >? ? 5392 ? ?3224 ? ?5984 ? ? ? 6100 ? ?5 1003 23:49:06 >? ? 1452 ? ?5240 ? ?5984 ? ? ? 8104 ? ?5 1003 23:49:06 >? ? 5240 ? ?3224 ? ?5984 ? ? ? 4532 ? ?5 1003 23:49:06 >... Defunct processes are not necessarily indicative of a cygwin problem. This could easily be a problem with gnuplot. cgf -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple