PR #2109 provides a solution for a race in directory creation (if a file
is to be written to a new path). There is a small weakness in that patch
in that it introduces a (very unlikely) potential hang condition during
shutdown.
see also https://github.com/rsyslog/rsyslog/pull/2109
against other processes working on the same directories, the repeated clashes are
most likely to occur within our address space. Seen to lead to failure loops with
rsyslogd starting two threads which continuously fail here until the service is
restarted manually.
Remove the use sizeof(char) or sizeof(uchar) in calculations for memory
allocation or string length. There are no known platforms for which
sizeof(char) or sizeof(uchar) is not 1, and c99 has defined sizeof(char)
to be 1 (section 6.5.3.4 of c99).
The affected code is used to detect a race condition in between
testing for the existence of a directory and creating it if it didn't
exist. The variable tracking the number of attempts wasn't reset for
subsequent elements in the path, thus limiting the number of
reattempts to one per the whole path, instead of one per each path
element.
This solution was provided by Martin Poole.
If multiple files try to create a directory at (almost) the same time,
some of them may fail. This is a data race and also exists with other
processes that may create the same directory. We do now check for this
condition and gracefully handle it.
This was not honored by the new ompipe module, because it is a local
file directive (it was applied to pipes as a side-effect of using the
same module for pipes and files...). I now made this a global, so that
semantics are the same as previously. Not really nice, but probably
the best thing to do in the current situation (everything else would
involve much more overhead --- leave that for the new config system).
support for enhancing probability of memory addressing failure by
using non-NULL default value for malloced memory (optional, only if
requested by configure option). This helps to track down some
otherwise undetected issues within the testbench and is expected
to be very useful in the future.
- bugfix: subtle (and usually irrelevant) issue in timout processing
timeout could be one second too early if nanoseconds wrapped
- set a more sensible timeout for shutdow, now 1.5 seconds to complete
processing (this also removes those cases where the shutdown message
was not written because the termination happened before it)
Unfortunatley, I do not have the full list of contributors
available. The patch set was compiled by Ben Taylor, and I made
some further changes to adopt it to the news rsyslog branch. Others
provided much of the base work, but I can not find the names of the
original authors. If you happen to be one of them, please let me
know so that I can give proper credits.
- added the ability to specify an error log function for the
runtime
- removed dependency of core runtime on dirty.h
Note that it is "better" modularity, not perfect. There is still
work to do, but I think we can for the time being proceed with
other things.
