Re: [boost] Boost.Log crash - question on consequences of misconfiguration

Subject: Re: [boost] Boost.Log crash - question on consequences of misconfiguration
From: Andrey Semashev (andrey.semashev_at_[hidden])
Date: 2014-02-11 21:46:32

On Tuesday 11 February 2014 20:15:35 Pieter wrote:
> 
> My question regarding Boost.Log is the following: could anything been said
> on what to expect when asking Boost.Log to write to a non-existing directory
> on a read-only drive? Does it keep trying to write or give up when the
> directory cannot be created, and would the user be informed in such a case?

The expected behavior is that Boost.Log tries to create the directories and 
the file on the first logged record but fails with an exception (during 
boost::filesystem::create_directories() call). If the exception is suppressed, 
this would happen on every logged log record.

> Background: today I ran into a problem using Boost.Log. I configured both a
> Console and a TextFile backend (both autoflushing) but happened to configure
> the latter wrongly- setting the location of the file on a not-existing
> directory on a read-only drive. This caused a program crash under certain
> circumstances -  but no problems under other circumstances (related to the
> configuration of the application which should not be relevant here) - when
> it did crash, this happened after a few messages were logged. Crash
> happened shortly before program termination but before the main() function
> was terminated (for a while I thought I was seeing the problem described in
> the docs regarding the boost.filesystem related problem and it took me a
> while to learn this was not the case).

It's hard to diagnose the problem not having a backtrace or a code sample, but 
my initial guess is that it was probably not a crash but the application 
terminated due to an uncaught exception.