| Age | Commit message (Collapse) | Author |
|---|
|
Parallel_run.py would write a ton of status updates as it runs. To
make it a bit cleaner have it overwrite the run-count line each time
instead.
Test: manual
Change-Id: I9073810fc7ece9a7d25b4a97a4bd01414ad5efa8
|
|
Make parallel_run just wait for any remaining jobs to finish before
exiting. Attempting to exit early generally didn't work very well.
Also improve documentation a bit.
Test: ./art/tools/parallel_run.py -j80 runit.sh
Change-Id: Ie8b3f1b4ad9bbc03a04e6bdc8e8cea6140b85f7f
|
|
Sometimes when investigating a flake it is useful to add logging and
repeatedly run a command until the issue appears with the logging.
This can be tedious if the flake is rare. To make it easier I added a
simple parallel_run.py tool that can run a command in parallel
repeatedly until it fails and save the stdout from the failed run.
Usage:
% # Write a script that exits with 0 on failure and writes all output
% # to stdout.
% cat tst.sh
\#!/bin/bash
! (head -c32 /dev/urandom | hd | grep -E '\|.*a')
% # Run the script in parallel
% ./art/tools/parallel_run.py -j80 --out out.txt ./tst.sh
Temporary files in /tmp/tmp5013_ey7
1 runs
<snip>
31 runs
% # Examine output
% cat out.txt
00000000 6d e3 ac bf 1a 21 93 17 61 00 2a e0 8b f5 41 3b |m....!..a.*...A;|
Test: ./art/tools/parallel_run.py -j70 --out out.txt ./tst.sh
Bug: 72608560
Change-Id: I83877ae1ccde5e843da20c5bd5a4f24e2e07fb19
|