GitHub user jaikiran opened a pull request:
https://github.com/apache/ant/pull/52
[master] Fix BZ-58451 BZ-58833
The commit here fixes the issues reported in:
https://bz.apache.org/bugzilla/show_bug.cgi?id=58833
https://bz.apache.org/bugzilla/show_bug.cgi?id=58451
The `PumpStreamHandler` and the `StreamPumper` work together for
redirecting the streams (reading from inputstream and writing to a
outputstream). When the `StreamPumper` is asked to stop, it stops reading from
the input stream and goes on to finish any (non-blocking) reads and finally a
(potentially blocking) flush of whatever it has read so far. The
`PumpStreamHandler`, which initiates that `stop` is a bit too impatient (mainly
because of its inability to know what's going on in `StreamPumper` after the
`stop` was invoked on it) and starts interrupting the `StreamPumper`'s thread
very soon (as early as 200 milli seconds after stop). This triggers a bunch of
issues in `StreamPumper`, which either is in the middle of finishing the
non-blocking reads or in the middle of actually flushing out whatever it's held
on to and ultimately leads to a non-clean stop and thus corrupting/truncating
the output stream as noted in those issues.
The commit here introduces a way, through the `PostStopHandle`, for
`PumpStreamHandler` to be aware that upon `stop` the `StreamPumper` has
respected the call to `stop` and is doing certain post-stop finishing acts
before actually being considered `finished`. The `PumpStreamHandler` then
allows a bit of time for the post-stop tasks to finish before actually trying
to kill off the `StreamPumper` using thread interrupts, if the `StreamPumper`
didn't finish after that grace period.
The commit here intentionally does _not_ introduce a configurable "grace
period" for the post-stop activities and instead using a hard coded (maximum) 2
seconds for the following reasons:
* The 2 seconds is the _maxium_ amount of time to let the post-stop to
complete, so it's not always going to wait that long
* The cleaning up gracefully is an internal implementation detail and
doesn't have to end up being configured by the user.
* Finally, the `PumpStreamHandler` anyway falls back to the thread
interrupts if the post-stop (flush() etc...) doesn't complete in those 2
seconds. So from a end user point of view, there isn't really any behaviour
change, except that we now give the `StreamPumper` a chance to gracefully stop.
This PR is against master branch. If this is approved I'll backport it to
1.9.x branch too.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/jaikiran/ant bz-58451-master
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/ant/pull/52.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #52
----
commit 3bcdac90581cb78e4a49b0b7fe6b01657a6436f1
Author: Jaikiran Pai <jaiki...@apache.org>
Date: 2017-12-16T09:48:30Z
BZ-58451 BZ-58833 Give StreamPumper a chance to finish cleanly before
interrupting its thread, to prevent truncated output
----
---
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@ant.apache.org
For additional commands, e-mail: dev-h...@ant.apache.org
