Peter Stamfest [Wed, 19 Feb 2014 08:56:16 +0000 (09:56 +0100)]
Add support for rrd restore to explicitly end xml files when using
the pipe interface by using the "-" special filename and using
ctrl-Z/newline as an end-of-file indicator.
Peter A. Bigot [Wed, 19 Feb 2014 00:31:01 +0000 (18:31 -0600)]
autogen.sh: use POSIX comparison for test
Although bash recognizes ==, and the shell has a shebang to use it, some
of us still automatically type "sh autogen.sh" instead of
"./autogen.sh", and this breaks on debian/jessie where /bin/sh is not
bash. The fix is trivial and should be safe for bash too (which
recommends using = instead of == with the test command for POSIX
conformance.
Peter A. Bigot [Mon, 17 Feb 2014 23:24:25 +0000 (17:24 -0600)]
rrd_graph: initialize new data member (fix #441)
primary_axis_format was added to image_desc_t but rrd_graph_init was not
updated to initialize it, resulting in this error when the garbage
values were used as axis labels:
(process:24893): Pango-WARNING **: Invalid UTF-8 string passed to pango_layout_set_text()
Jan Kundrát [Wed, 26 Feb 2014 01:57:19 +0000 (02:57 +0100)]
Fix rrdc_fetch behavior on RRDs with multiple DSes
When reading values from RRD files with multiple data sets, the auxiliary
parsing function would only copy the first item from the returned list; the
others were left uninitialized. On my x64_64 Linux machine, this actually reused
values form previous call to parse_char_array_header which happens to be a DS
definition. Given a RRD file with DS "a", "b" and "c", the first value on each
line would be correctly parsed, but the second value is "b", the actual DS name,
and the third one "c". These "values" cannot be parsed as numbers, and the whole
function therefore returns with an error. Making sure that all pointers within
the array are copied therefore fixes this problem.
Markus Elfring [Mon, 25 Nov 2013 16:16:25 +0000 (17:16 +0100)]
Bug #429: Clean-up of reserved identifier violations
A few identifiers did not fit to the expected naming convention of the
C++ language standard. This detail was fixed by the deletion of an underscore.
The probability for name clashes was reduced for the affected include guards
by the addition of a kind of universally unique identifier as a suffix.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Markus Elfring [Mon, 25 Nov 2013 16:10:51 +0000 (17:10 +0100)]
Bug #429: Clean-up of reserved identifier violations
Some identifiers did not fit to the expected naming convention of the
C language standard. This detail was fixed by the deletion of leading underscores.
The probability for name clashes was reduced for the affected include guards
by the addition of a kind of universally unique identifier as a suffix.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Markus Elfring [Mon, 25 Nov 2013 13:11:34 +0000 (14:11 +0100)]
Bug #431: Fix signal handler
The signal handler implementation called async-signal-unsafe functions.
Corresponding notifications are handled by the interface "sigwaitinfo"
in a separate thread now.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Steven Hartland [Mon, 7 Oct 2013 11:21:37 +0000 (11:21 +0000)]
== Makefile / config fixes ==
* Fixed invalid AC_LANG_PROGRAM usage.
* Fixed AUTOMAKE_OPTIONS not defining subdir-objects which caused warnings
* Switched from legacy INCLUDES to AM_CPPFLAGS
* Added Makefile clean targets to ensure all autogenerated targets are removed
* Updated Makefile.PL to be more portable and less hardcoded for Win32
* Switched bash scripts to use #!/usr/bin/env bash for compatibility with none Linux OS's
* Cleaned up win32/Makefile.msc variable defines for easy updating
* Updated defines of SIZEOF_TIME_T to take into account the variances in Windows.
== Code Enhancements / fixes ==
* Added support for rrdcached client methods to windows builds
* Fixed invalid type warning for atoi assignment to a enum
* Fixed malloc cast warnings
* Fixed illegal use of this for variable
* Fixed getsockname cast warning
* Removed HAVE_CONFIG_H checks as all platforms now have it
* Removed unused skip variable from rrd_graph data_fetch
* Fixed unused var warnings for newGraphDescription params
* Fixed unused var st warning in HandleInputLine
* Fixed all includes for <stdint.h> to check HAVE_STDINT_H
* Fixed use of uninitialsed base_dir in rrd_mkdir_p on Win32
* Updated win32/README with details of obtaining stdint.h as well as locations
for dependencies
Christoph Biedl [Sun, 6 Oct 2013 21:41:27 +0000 (23:41 +0200)]
Add a --legend-direction=bottomup variant
The --legend-direction=bottomup2 variant will exempt leading and
trailing COMMENT lines from re-ordering. This might be useful for
generators like munin that use such lines for table headers and the
like.
Julien Thomas [Tue, 25 Jun 2013 11:22:40 +0000 (13:22 +0200)]
Allow empty imginfo string format
Commit 37d3050 "Added imginfo format check" breaks -f "" which used
to be very usefull to print data as text format on stdout. Allowing
an empty -f was particularly interesting because there was no
<width>x<height> printed out.
Julien Thomas [Tue, 25 Jun 2013 11:11:27 +0000 (13:11 +0200)]
Add prototype for bad_format_imginfo
This patch solve implicit declaration of function 'bad_format_imginfo'
warning introduced by commit 37d3050 "Added imginfo format check".
rrd_graph.c: In function 'rrd_graph_v':
rrd_graph.c:4025: warning: implicit declaration of function 'bad_format_imginfo'
rrd_graph.c:4025: warning: nested extern declaration of 'bad_format_imginfo'
rrd_graph.c: At top level:
rrd_graph.c:4836: warning: no previous prototype for 'bad_format_imginfo'
Vincent Bernat [Mon, 8 Apr 2013 16:16:49 +0000 (18:16 +0200)]
graph: when an RRD file is missing, provide an empty one
Instead of providing a RRD without any DS, we provide one with the
expected DS. This allows one to use CDEF with this RRD. The DS
contains only NaN values.