runProcess: misbehaving exception on nonexistent working directory

[wferi]

Compiling and running the attached Fork.hs gives unexpected results:

• The first line is mostly OK: the working directory does not exist. (Unfortunately it's not clear from the message what does not exist...)
• The ensuing ps output is also OK, but already shows strange process states and hierarchy:
  • Why is the parent Fork in D state?
  • Why is ps a grandchild of it, instead of being another child of the parent Fork?
• The last line is Segmentation fault, which is bad.

$ ./Fork
sleep: runProcess: does not exist (No such file or directory)
  PID TTY      STAT   TIME COMMAND
 3504 ?        Ss     0:00 urxvt
 3505 ?        S      0:00  \_ urxvt
 3506 pts/1    Ss     0:00  \_ bash
 5571 pts/1    D+     0:00      \_ ./Fork
 5572 pts/1    S+     0:00          \_ ./Fork
 5573 pts/1    R+     0:00              \_ ps xf
Segmentation fault

(I elided the irrelevant part of the ps output.)

Somebody on IRC got Bus error instead. Others, testing under GHC 6.8, did not even get an exception. I have got the impression that exception handling inside or around runProcess is broken.

[igloo]

Thanks for the report. Not sure we'll get to it for 6.8.2, but we should try to have a look as segfaults are bad!

[simonmar]

mine

[simonmar]

dup of #1593, fixed in 6.8.1