---
- branch: netbsd-8
  date: Sat Aug 25 11:45:40 UTC 2018
  files:
  - new: 1.140.2.4
    old: 1.140.2.3
    path: src/bin/sh/eval.c
    pathrev: src/bin/sh/eval.c@1.140.2.4
    type: modified
  id: 20180825T114540Z.8fb7a62fc0c12be4283fed3738bb156917d86255
  log: "Pull up following revision(s) (requested by kre in ticket #982):\n\n\tbin/sh/eval.c:
    revision 1.157\n\nPR bin/42184 PR bin/52687  (detailing the same bug).\n\nFix
    \"command not found\" handling so that the error message\ngoes to stderr (after
    any redirections are applied).\n\nMore importantly, in\n\n\tfoo > /tmp/junk\n\n/tmp/junk
    should be created, before any attempt is made\nto execute (the assumed non-existing)
    \"foo\".\n\nAll this was always true for any command (not found command)\ncontaining
    a / in its name\n\n\tfoo/bar >/tmp/junk  2>>/tmp/errs\n\nwould have created /tmp/junk,
    then complained (in /tmp/errs)\nabout foo/bar not being found.   Now that happens
    for ordinary\ncommands as well.\n\nThe fix (which I found when I saw differences
    between our\ncode and FreeBSD's, where, for the benefit of PR 42184,\nthis has
    been fixed, sometime in the past 9 years) is\nfrighteningly simple.   Simply do
    not short circuit execution\n(or print any error) when the initial lookup fails
    to\nfind the command - it will fail anyway when we actually\ntry running it.   The
    cost is a (seemingly unnecessary,\nexcept that it really is) fork in this case.\n\nThis
    is what I had been planning, but I expected it would\nbe much more difficult than
    it turned out....\n\nXXX pullup-8\n"
  module: src
  subject: 'CVS commit: [netbsd-8] src/bin/sh'
  unixtime: '1535197540'
  user: martin