3.5/f/cfg tree final fix #26
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
On Linux 3.2, /dev/kmsg is present, is seekable, is pollable with poll(), but not with epoll(). This leads to a run-time abort, because syslog-ng can't monitor the file in this case. To fix the situation, tighten the requirements in system-source, and check not only for seekability, but pollability too. Reported-by: Martin Bagge / brother <brother@bsnet.se> Signed-off-by: Gergely Nagy <algernon@balabit.hu>
Using /dev/pipe on HP-UX was a mistake, use /dev/log instead, as the original shell script did. Signed-off-by: Gergely Nagy <algernon@balabit.hu>
It should look for files under both ${top_srcdir} and ${top_builddir},
not twice at the same place.
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
The afstreams module has been broken for more than a year, this patch brings it back from the dead by migrating it to newer APIs and all. Based on a patch by Balazs Scheidler <bazsi@balabit.hu>. Signed-off-by: Gergely Nagy <algernon@balabit.hu>
On Solaris, uname() can return any positive value, and that indicates success. Pretty much everywhere else, only 0 does. Thankfully, -1 signals failure everywhere, so check for that. This makes system() work on Solaris again. Signed-off-by: Gergely Nagy <algernon@balabit.hu> Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
When syslog-ng creates a pipe, create it with user-specified permissions or restrictive ones (0600) by default, instead of creating it with broad read-write permissions (0666) and patching it up later. Signed-off-by: Budai Laszlo <lbudai@balabit.hu> Signed-off-by: Gergely Nagy <algernon@balabit.hu>
On HPUX, during boot the console does not considered as a tty, so syslog-ng does not call setsid, the tty is not detached and the supervisor is killed at the end of the boot. setsid() does not and should not depend on isatty(), and anyways setsid() is considered more unix way than ioctl(). Signed-off-by: Juhasz Viktor <jviktor@balabit.hu> Signed-off-by: Viktor Tusa <tusa@balabit.hu>
Signed-off-by: Daniel Gados <gdani@balabit.hu>
The old key was leaked when renaming a persist-state key. This patch fixes that leak. Signed-off-by: Juhasz Viktor <jviktor@balabit.hu> Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
…n persist_state_load_v4 Signed-off-by: Viktor Juhasz <jviktor@balabit.hu> Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
When stats-level() is changed and syslog-ng is reloaded, an assertion failed in the facility counters unregistration. The reason was that the unregistration used the 'other' category instead of the one assigned to the given facility code. Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
Update the Travis-CI control file for the new branch structure. Signed-off-by: Gergely Nagy <algernon@balabit.hu>
The Linux kernel used to have a bug that caused the afuser driver to block even if O_NONBLOCK is set in the open flags. It was fixed a long time ago, and alarm() is incompatible with multithreaded programs, thus get rid off the alarm() call to detect blocked ttys. Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
Rearrange the README a little bit, putting the source installation docs at the top, adding more links, and rewording a few things here and there. Signed-off-by: Gergely Nagy <algernon@balabit.hu>
When building test programs (anything in ${check_PROGRAMS}) add
-no-install to the list of LDFLAGS, so that libtool stops generating a
useless wrapper for them. This makes debugging the check programs
easier, and also avoids a lot of shell launching for no good reason.
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
On 32-bit platforms, explicitly cast struct timeval members to 64-bit, so the time difference calculations will not be truncated. Reported-by: Jakub Wilk <jwilk@debian.org> Signed-off-by: Gergely Nagy <algernon@balabit.hu>
The host() and port() options were not declared in the parser, only in the grammar, thus using them led to a syntax error. Declare them for the parser too. Reported-by: Otto Berger <otto@bergerdata.de> Signed-off-by: Gergely Nagy <algernon@balabit.hu>
When the message is not local, do not override - or set - the value of PROGRAM, there's other ways to do that. Doing an unconditional setting breaks setups where kernel messages are forwarded between syslog-ng instances: The first syslog-ng sees "<4>Some message..." on /proc/kmsg, sets PROGRAM to "kernel", and sends it over on tcp as "<4>kernel: Some message...". The second syslog-ng sees that, from the facility it sees that it's a kernel message, and sets PROGRAM to kernel again. But when a message is from the kernel, we don't parse the rest, we don't extract the program, we treat the rest of the message as MESSAGE, so in this case, we'd end up with "<4>kernel: kernel: Some message", and each hop would add another "kernel: ". If we only set this when the message is local, we avoid this. The downside is, that on the remote end, PROGRAM will not be set. Reported-by: Matyas Koszik <koszik@atw.hu> Signed-off-by: Gergely Nagy <algernon@balabit.hu>
Since OpenSSL needs some initialization to work properly, link afql to libsyslog-ng-crypto to do that. If we don't do this, then loading afsql (or any other module that uses OpenSSL) before anything that uses libsyslog-ng-crypto (like afsocket) will fail and result in a segmentation fault. Affected modules are: afsql, afmongodb and afsmtp. Reported-by: Imre Lazar <imre@balabit.hu> Signed-off-by: Gergely Nagy <algernon@balabit.hu>
NEWS & VERSION updated for 3.5.2. Signed-off-by: Gergely Nagy <algernon@balabit.hu>
…to a destination This patch adds validation that no extra messages get delivered to a destination. Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
This patch fixes the application of node flags during compilation. Earlier the node flags were either applied to the first_pipe (normal log paths lacking a source reference) or the last_pipe (log paths that did contain source references). The issue is that due to the compilation of junctions, the last_pipe may not be reachable through LogPipe->pipe_next list, which is iterated in log_multiplexer_init() when looking for LC_XXX flags. Rather they are only reachable through an intermediate LogMultiplexer's next_hop list. This causes the flags to be missed when evaluating the compiled structure. This patch tries to apply flags to the first LogPipe ever encountered while compiling the sequence and only resort to the last_pipe if we don't encounter anything that would start with a sub_pipe_head. Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
…ments The final flag was incorrectly processed for log statements where it was specified after a set of embedded log statements. This adds a testcase that validates that. Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
|
Yep, will pick it for those two tomorrow. Thanks for the notice! |
|
Merged it to 3.5, and cherry-picked the patches to 3.4, thanks! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This should fix a config compilation problem, and affects both 3.4 and 3.5, and probably be fixed in both.
@algernon: could you please apply this to both 3.4 and 3.5? The bug was reported on the mailing list two weeks ago with the subject: [syslog-ng] syslog-ng 3.5.1 - question about flags(final)...
I'm integrating the same fix to 3.6/master. Thanks.