[Elvand-Lie]

This Pull request:

Changes or fixes:

A range set only on the Z axis of a TProfile3D is ignored by GetMean and GetStdDev:

TProfile3D p("p", "", 4, 0, 4, 5, 0, 10, 3, 0, 30);
p.Fill(0.5, 1., 5., 1.);
p.Fill(3.5, 9., 25., 1.);
p.GetZaxis()->SetRange(1, 1);
// master: GetMean(1) == 2, GetMean(3) == 15 (both points still counted)
// TH3D with the same fills and range: GetMean(1) == 0.5, GetMean(3) == 5

TProfile3D::GetStats recomputes the statistics from the bins in range only when the X or Y axis has a range:

if ((fTsumw == 0) || fXaxis.TestBit(TAxis::kAxisRange) || fYaxis.TestBit(TAxis::kAxisRange)) {

TH3::GetStats also checks the Z axis, so this change adds fZaxis.TestBit(TAxis::kAxisRange) to the condition.

Taking that path for a Z range exposed a second problem. The recomputation loop never includes underflow and overflow bins, even when overflow statistics are enabled (SetStatOverflows(TH1::kConsider) or TH1::StatOverflows(true)). With a Z range, an entry in the X underflow was then dropped. The loop now uses the per-axis bounds of TProfile2D::GetStats: on an axis without a range, the underflow and overflow bins are included when overflow statistics are enabled.

Two tests in hist/hist/test/test_TProfile.cxx:

• TProfile3D.GetStatsHonoursZRange: the example above; the means are the centers of the one bin in range.

• TProfile3D.GetStatsHonoursStatOverflowsWithZRange: with SetStatOverflows(TH1::kConsider) and a Z range, an entry in the X underflow is still counted (sum of weights 2, X mean 0, as with TH3D).

• Both tests fail without the corresponding change and pass with it.

• ctest -R '^gtest-hist-hist-' passes: 33 tests, 0 failures, on Linux with a minimal build.

• Also checked against TH3D: every combination of X, Y and Z ranges with each overflow-statistics setting, all means and standard deviations.

• Not run: the full ROOT test suite, roottest, and platforms other than Linux.

Checklist:

• tested changes locally
• updated the docs (if necessary)

Antigravity (Claude Opus 5.5) and Codex (gpt-6.1-sol) implemented the fix and the tests; Claude Code (Claude Opus 5.5) found the bug and validated it.