On Wed, Feb 25, 2004 at 01:57:37PM -0500, Rolf Campbell wrote: >Christopher Faylor wrote: >>>>--- t.sh --- >>>>#!/bin/bash >>>>export C=1 >>>>while make -j ; do C=$(($C+1)) ; done >>>>echo Failed after $C runs 1>&2 >>>>--- end of t.sh --- >>>> >>>>The script failed with: >>>> >>>>$ ./t.sh > freeze.out >>>>/bin/sh: line 1: sleep: No such file or directory >>>>make: *** [12.pp] Error 127 >>>>make: *** Waiting for unfinished jobs.... >>>>Failed after 1499 runs >>> >>>I ran a variation of the above for three days without fail so I think I >>>can safely say that I can't reproduce this problem. >> >>And, by the way: http://cygwin.com/ml/cygwin/2004-02/msg00923.html > >I tried running that script again, here were my results: > >Feb14: more than 10,000 iterations (never failed, just got bored of >watching it) >Feb17: more than 270 (still running) >Feb18: Froze after 12, 41, 6 >Feb20: Froze after 9, 2, 4 >Feb21: Froze after 1, 5, 4
Out of curiousity, the next time it freezes, try hitting CTRL-D. I ran one of the scripts for four hours and got a freeze. gdb showed that make was sitting in console input mode for some reason. Hitting ctrl-d (followed two seconds by an expletive) got things running again. I didn't really expect that to "fix" things, though. I really wanted to poke around to see why it was waiting for input. So, if this is correct then putting a "< /dev/null" in the script would probably make things run forever, not that that should be the ultimate solution. >My un-educated guess is: >2004-02-17 Corinna Vinschen <[EMAIL PROTECTED]> > > * fork.cc (fork_child): Move fixup_shms_after_fork so that > signal_arrived is initialized when calling it. > It is unlikely in the extreme that this would be the cause. make doesn't use shm, AFAIK. cgf -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Problem reports: http://cygwin.com/problems.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
