RRDtool Version 1.11.0

RRDtool 1.11.0 - 2026-08-07

Bugfixes

• hardening: size the rrd_diff scratch buffer for its largest input @somethingwithproof
• rrdtool dump now uses a numeric UTC offset (%z, e.g. +0900) instead
  of the locale timezone name (%Z) in XML comments. Locale timezone names
  can be encoded in a non-UTF-8 encoding (e.g. CJK locales), which breaks
  the XML declaration and causes rrdtool restore to fail with
  "Input is not proper UTF-8". The numeric offset is always ASCII.
  Note: existing dump files already containing locale timezone names remain
  valid XML as long as their encoding matches; newly created dumps will
  always contain pure ASCII in these comments. Issue #1173
  reported by @k79e, fix by @oetiker.
• Fix segfault when --x-grid is given without a trailing date-format
  field. The %n conversion in sscanf() is not counted in its return
  value, so stroff remained uninitialized when the input ended before
  the final : separator; the subsequent stroff != 0 guard read
  garbage and could branch into strdup() with a bogus offset. Fixed by
  initialising stroff = 0 so a partial match safely falls through to
  the existing "invalid x-grid format" error path. Issue #1291
  reported by @anvilvapre, fixed by @oetiker
• libdbi /derive now returns 0 instead of NaN when the underlying
  counter is constant (zero delta). Previously the strict d_value > 0
  guard skipped zero deltas the same way it skips negative ones (counter
  resets). Changed to d_value >= 0. Issue #550 reported by
  @bitionaire, fix by @oetiker.
• Fix a 1-byte heap buffer overflow in checkUnusedValues() that
  triggered *** buffer overflow detected *** SIGABRT under glibc
  fortified strcat() when reporting unused graph arguments. The
  function under-allocated by one byte and silently relied on that
  overflow to make its trailing : strip land on the right index.
  Refactored to size buffers correctly and prepend the separator
  rather than build-then-strip, so the two no longer have to agree.
  Reported and original fix attempt in PR #1329 by @ppisar,
  refactored fix by @oetiker
• Pad the Perl $RRDs::VERSION / $RRDp::VERSION numeric encoding so
  two-digit minor releases compare monotonically. The numeric version
  now uses three-digit zero-padded minor and patch fields, e.g.
  1.10.3 -> 1.010003 (previously 1.10003, which was numerically less
  than 1.9.0 -> 1.9000). The same encoding is now used by NUMVERS
  in configure.ac and win32/rrd_config.h. Note: consumers that
  previously tested $RRDs::VERSION >= 1.2 need to update their
  checks (use >= 1.002 or compare against the X.Y.Z string via the
  version pragma) -- this is unavoidable given the existing format
  could not represent two-digit minors monotonically. Issue #1330
  reported by @ppisar, fixed by @oetiker
• Y-axis labels keep enough fractional precision when the gridstep is
  fractional after SI scaling (e.g. values around 10G+), avoiding
  duplicate labels on tall graphs. Issue #1326 reported by @muusik,
  fixed by @oetiker
• Detect nroff support for -Wbreak at configure time, so building the
  text documentation works on systems with BSD-derived nroff (e.g.
  macOS /usr/bin/nroff). Issue #1323 reported by @gigaimage, fixed by
  @oetiker
• Fix SIGFPE divide-by-zero crash in rrdtool dump when processing
  fuzz-crafted RRD files that contain an RRA with pdp_cnt or
  pdp_step equal to zero. The modulo operation at the heart of the
  timestamp calculation is now guarded: such files are rejected with
  a descriptive error message instead of crashing.
  Issue #783 reported by @rwhitworth, fix by @oetiker.

Features

• rrdgraph: new --vertical-label-angle and --right-axis-label-angle options
  to rotate the two y axis legends independently at run time. The compile
  time RRDGRAPH_YLEGEND_ANGLE remains the default for both, so existing
  graphs are unchanged. Use --right-axis-label-angle 270 to have the right
  hand legend read from the top down. Suggested by @blappm in #1377,
  implemented by @oetiker
• rrdgraph: new --right-axis-range min:max option to suppress right axis
  labels outside the given range, so a bounded right axis is no longer
  labelled with values it cannot reach when the left axis range does not
  line up. Either end may be left empty for unbounded. Suggested by @blappm
  in #1377, implemented by @oetiker