Re: [PATCH v9 00/19] index-helper/watchman

On 10/05/16 00:12, David Turner wrote:

On Mon, 2016-05-09 at 15:32 -0700, Junio C Hamano wrote:

quoted

Junio C Hamano [off-list ref] writes:

quoted

David Turner [off-list ref] writes:

quoted

On Mon, 2016-05-09 at 14:40 -0700, Junio C Hamano wrote:

quoted

Hmmm, I seem to be getting

    $ cat t/trash*7900*/err
    fatal: Already running

after running t7900 and it fails at #5, after applying
"index-helper: optionally automatically run"

The symptom looks pretty similar to $gmane/293461 reported earlier.
Here is how "t7900-index-helper.sh -i -v -x -d" ends.


expecting success:
        test_when_finished "git index-helper --kill" &&
        rm -f .git/index-helper.sock &&
        git status &&
        test_path_is_missing .git/index-helper.sock &&
        test_config indexhelper.autorun true &&
        git status &&
        test -S .git/index-helper.sock &&
        git status 2>err &&
        test -S .git/index-helper.sock &&
        test_must_be_empty err &&
        git index-helper --kill &&
        test_config indexhelper.autorun false &&
        git status &&
        test_path_is_missing .git/index-helper.sock

+ test_when_finished git index-helper --kill
+ test 0 = 0
+ test_cleanup={ git index-helper --kill
                } && (exit "$eval_ret"); eval_ret=$?; :
+ rm -f .git/index-helper.sock
+ git status
On branch master
Untracked files:
  (use "git add <file>..." to include in what will be committed)

        err

nothing added to commit but untracked files present (use "git add" to
track)
+ test_path_is_missing .git/index-helper.sock
+ test -e .git/index-helper.sock
+ test_config indexhelper.autorun true
+ config_dir=
+ test indexhelper.autorun = -C
+ test_when_finished test_unconfig  'indexhelper.autorun'
+ test 0 = 0
+ test_cleanup={ test_unconfig  'indexhelper.autorun'
                } && (exit "$eval_ret"); eval_ret=$?; { git index
-helper --kill
                } && (exit "$eval_ret"); eval_ret=$?; :
+ git config indexhelper.autorun true
+ git status
error: last command exited with $?=141

I think that's a SIGPIPE on the first git status.  Weird, since I just
added sigpipe-avoidance code (in v8).  Does anyone have any idea why
the sigchain stuff isn't doing what I think it is?

Sorry for a late report (I've been a bit busy last couple of days), but
I've been seeing exactly the same on v8 of this series.

Note that the above 'git status' is actually the second git-status in the
test.

I haven't been able to debug it too much, but I can tell you that it is
not failing at exactly the same place every time (so it may be time
sensitive). However, it often fails in poke_and_wait_for_reply() at the
first packet_flush() (which in turn calls write_or_die()  which calls
check_pipe() with an EPIPE(32)). At other times it fails when issuing
a flush after a refresh packet. For example, on one run with packet
tracing enabled, I got this for the trace:

  trace: built-in: git 'status'
  packet:          git> poke 3221 
  packet:          git> 0000
  packet:          git< OK
  packet:          git> refresh
  packet:          git> 0000

So, its getting the EPIPE for the refresh in this case, even though
the index-helper is still running, the unix socket is in .git/
(and so is a shm-* file BTW).

I didn't get any further than that I'm afraid.

ATB,
Ramsay Jones