[
https://issues.apache.org/jira/browse/LOG4NET-501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16322929#comment-16322929
]
Matthias Hess edited comment on LOG4NET-501 at 1/11/18 8:54 PM:
----------------------------------------------------------------
Hi, I guess my point was this:
{quote}The finalizers of two objects are not guaranteed to run in any specific
order, even if one object refers to the other. That is, if Object A has a
reference to Object B and both have finalizers, Object B might have already
been finalized when the finalizer of Object A starts.
{quote}
taken from https://msdn.microsoft.com/en-us/library/system.object.finalize.aspx
I can't remember this issue exactly and I haven't looked into the code today,
but I assume that RollingFileAppender has a Finalizer just like the FileStream
has one. RollingFileAppender (Object A) references the FileStream (Object B)
directly or indirectly. Finalization happens in undefined order: in my observed
case, apparently the FileStream was finalized *before* the RollingFileAppender.
Hence, RollingFileAppender failed to write a footer and flush the FileStream in
its own finalizer implementation.
was (Author: m.hess):
Hi, I guess my point was this:
> The finalizers of two objects are not guaranteed to run in any specific
> order, even if one object refers to the other. That is, if Object A has a
> reference to Object B and both have finalizers, Object B might have already
> been finalized when the finalizer of Object A starts.
taken from https://msdn.microsoft.com/en-us/library/system.object.finalize.aspx
I can't remember this issue exactly and I haven't looked into the code today,
but I assume that RollingFileAppender has a Finalizer just like the FileStream
has one. RollingFileAppender (Object A) references the FileStream (Object B)
directly or indirectly. Finalization happens in undefined order: in my observed
case, apparently the FileStream was finalized *before* the RollingFileAppender.
Hence, RollingFileAppender failed to write a footer and flush the FileStream in
its own finalizer implementation.
> Finalization of appenders might fail
> ------------------------------------
>
> Key: LOG4NET-501
> URL: https://issues.apache.org/jira/browse/LOG4NET-501
> Project: Log4net
> Issue Type: Bug
> Components: Core
> Affects Versions: 1.2.10
> Reporter: Matthias Hess
>
> Given the following example:
> {code}
> LogLog.InternalDebugging = true;
> RollingFileAppender result = new RollingFileAppender();
> result.File = ...;
> result.AppendToFile = true;
> result.Encoding = Encoding.UTF8;
> ...
> result.ActivateOptions();
> {code}
> When this rolling file appender object gets finalized (due to no more
> references), the finalization code tries to close the underlying file stream.
> This will fail, if the underlying file stream itself was already subject to
> finalization.
> It gives me log4net internal error messages when using the appenders outside
> of a logger repository.
> I think, the current finalization code does a little too much. Writing a
> footer/flushing should be triggered by the AppDomain events only.
> What do you think about it?
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
