[
https://issues.apache.org/jira/browse/MESOS-1870?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14170683#comment-14170683
]
Adam B commented on MESOS-1870:
-------------------------------
commit 3eb585981b95801eae174a8ff52d581f20b09247
Author: Cody Maloney <[email protected]>
Date: Tue Oct 14 00:24:26 2014 -0700
libprocess: Always use stout ABORT() rather than system abort()
This makes it so any time there is an abort, we get a line number and at
least
a basic message as to why there was an abort. If you want a clean(er) exit,
use <stout/exit>.
Review: https://reviews.apache.org/r/26473
commit c713c5d194a43006a788ef6217753f10845949b0
Author: Cody Maloney <[email protected]>
Date: Tue Oct 14 00:21:55 2014 -0700
stout: Always use stout ABORT() rather than system abort()
This makes it so any time there is an abort, we get a line number and at
least
a basic message as to why there was an abort. If you want a clean(er) exit,
use
<stout/exit>.
Also adds an overload which takes a standard string and unwraps it to a
const char * automatically, since a lot of the time we are building strings
to pass them to abort.
Review: https://reviews.apache.org/r/26472
> stout ABORT() should print a backtrace
> --------------------------------------
>
> Key: MESOS-1870
> URL: https://issues.apache.org/jira/browse/MESOS-1870
> Project: Mesos
> Issue Type: Bug
> Reporter: Cody Maloney
> Priority: Minor
>
> Currently we die fast and hard with little (if any) details about why we
> died. This gets particularly hard to debug around Try<>::get(), which uses a
> raw OS abort, and could be called from arbitrary surrounding code.
> Anything that is immediate process termination which we don't know to have a
> really nice, clean error message, should call stout ABORT(), which ideally
> will print a backtrace to aid with debugging.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
