It appears that I have discovered a new fork() of death.  :-(  This one
results in system abort 614 (unexpected ESCAPE).

On Monday, I installed GNUCORE and GNUGCC from the IPROF98 Freeware tape.
Since then, I've crashed my porting machine THREE times while doing full
recompiles of sendmail 8.9.0Beta5.

Dumps were taken, an HPRC call was opened, HPRC read the dumps, and
SR 5003-414110 was entered because this appears to be a new unknown problem.

Here's the stack trace of the process that crashed the system:

$14f ($66) nmdat > pin 5b
$150 ($5b) nmdat > tr
     PC=a.0018604c system_abort
* 0) SP=4183f0d0 RP=a.00f1ef24 fs_duplicate+$4d4
  1) SP=4183f0d0 RP=a.00c36a58 fs_setup_resv_files.fs_open_sys_file+$244
  2) SP=4183ed90 RP=a.00c3720c fs_setup_resv_files+$5ac
  3) SP=4183ecd0 RP=a.0052a188 pm_initiate+$104
  4) SP=4183e810 RP=a.00529b44 outer_block+$a0
  5) SP=4183e610 RP=a.00000000
     (end of NM stack)

This is a child process still being fork()ed; it is attempting to duplicate
the parent's open file environment.

Here is the session process tree:

$149 ($5b) nmdat > dptree
...deletia...
  2a (/SYS/PUB/JSMAIN)
   57 (/SYS/PUB/CI)
    5d (/SYS/HPBIN/SH)
     5c (/usr/local/bin/make)
      5e (/SYS/HPBIN/SH)
       62 (/SYS/HPBIN/SH)
        66 (/SYS/HPBIN/SH)
         5b ()

What I had done to get to that point in the tree:

        :xeq sh.hpbin.sys -L
        cd /SENDMAIL/PUB/sendmail-890b5-m
        make

It successfully compiles everything in src/, but when it gets to mail.local/,
KA-BOOM!

Here is the parent's stack trace:

$14e ($66) nmdat > tr
     PC=a.0017b70c enable_int+$2c
* 0) SP=4183fcc8 RP=a.0147a0c0 notify_dispatcher.block_current_process+$324
  1) SP=4183fcc8 RP=a.0147c548 notify_dispatcher+$264
  2) SP=4183fc48 RP=a.001abc20 wait_for_active_port+$ec
  3) SP=4183fb48 RP=a.001ac8a0 receive_from_port+$534
  4) SP=4183fac8 RP=a.00349184 ipc_wait_process+$3a8
  5) SP=4183f8c8 RP=a.014ab17c create_a_process.start_new_process+$3a4
  6) SP=4183f688 RP=a.014abe88 create_a_process+$a70
  7) SP=4183f548 RP=a.0052632c ?create_a_process+$8
       export stub: a.0023e73c px_fork+$6d4
  8) SP=4183e448 RP=a.0023e054 ?px_fork+$8
       export stub: 174.011ac0d4 _px_fork+$18
  9) SP=4183e188 RP=174.011ac088 ?_px_fork+$8
       export stub: 39d.0004aca8 _fork+$1c
  a) SP=4183e108 RP=39d.0002a614 j_fork+$e8
  b) SP=4183e0c8 RP=39d.0001cdd0 e_cmd+$494
  c) SP=4183e048 RP=39d.0001b558 execute+$2f8
  d) SP=4183df08 RP=39d.0001b630 execute+$3d0
  e) SP=4183ddc8 RP=39d.0001b820 execute+$5c0
  f) SP=4183dc88 RP=39d.000181f0 shell+$224
 10) SP=4183db48 RP=39d.00017d00 mks_main+$9f4
 11) SP=4183dac8 RP=39d.0003e56c main+$80
 12) SP=4183d9c8 RP=39d.00044ecc _start+$bc
 13) SP=4183d8c8 RP=39d.000172ec $START$+$1c
 14) SP=4183c748 RP=39d.00000000
     (end of NM stack)

Fortunately this doesn't happen every time I try to recompile sendmail.
Unfortunately it has happened 3 times in the past 3 days.  :-(

Perhaps the HP people doing the Java compiles would like to install gcc 2.8
from the IPROF98 Freeware tape; I'm sure repeated SA614 crashes inside of HP
would help to get this problem fixed promptly.  ;-)
--
Mark Bixby                      E-mail: [log in to unmask]
Coast Community College Dist.   Web: http://www.cccd.edu/~markb/
District Information Services   1370 Adams Ave, Costa Mesa, CA, USA 92626-5429
Technical Support               Voice: +1 714 438-4647
"You can tune a file system, but you can't tune a fish." - tunefs(1M)