[aywrite]

The texel:1.08 rung of the calibration ladders stopped building when the runner image moved to cmake 4. Texel bundles a gtest for its tests whose cmake file asks for cmake 2.8.12, and cmake 4 refuses to configure anything asking for less than 3.5. In Calibrate runs 43527508936 and 43527512769 the rung played no games and was left out of the fit:

CMake Error at test/gtest/CMakeLists.txt:56 (cmake_minimum_required):
  Compatibility with CMake < 3.5 has been removed from CMake.

The configure step in scripts/opponent.sh now passes -DCMAKE_POLICY_VERSION_MINIMUM=3.5, which is what cmake's own message suggests. The repository, the pin and the build options are unchanged, so the rated release is still what plays. A cmake older than 4 ignores the variable.

Checked. Built locally with cmake 4.4.4 and 3.28.3; both binaries answer uci as Texel 1.08. The script tests pass (365). On the runner, Calibrate run 43553675050 from this branch built the opponent and is playing the rung.

---

Generated by Claude Code

[github-actions]

Speed against 3ab386ca

Measured on INTEL(R) XEON(R) PLATINUM 8573C.
Both sides built with rustc 1.99.1 (58cf8f9c3 2026-10-13).

Both sides built and run on this runner in this job, the way
scripts/speed.sh measures a perf commit. Each round runs
both sides on a layout of its own, the same compiled code with
its code and data shuffled and moved, so the 95% interval carries
where the code landed as well as the run. The verdict holds the
interval against a 1% threshold. The default layout, the one a
release ships, is measured after as a diagnostic. The node
counts are the search's: they move when the search does, and a
speed change leaves them alone.

round     base nps  candidate nps  change
    1      5490621        5491136   +0.0%
    2      5553567        5556320   +0.0%
    3      5527670        5491560   -0.7%
    4      5517893        5527745   +0.2%
    5      5480630        5481359   +0.0%
    6      5529718        5530132   +0.0%
    7      5520239        5524145   +0.1%
    8      5536112        5536293   +0.0%
    9      5518443        5546544   +0.5%
   10      5549155        5547602   -0.0%
   11      5539056        5555047   +0.3%
   12      5524263        5528687   +0.1%
   13      5485292        5500048   +0.3%
   14      5558380        5557945   -0.0%
   15      5528664        5527417   -0.0%
   16      5527516        5531413   +0.1%
   17      5483476        5487858   +0.1%
   18      5542177        5528205   -0.3%
   19      5563071        5556731   -0.1%
   20      5545309        5544203   -0.0%
   21      5502677        5492373   -0.2%
   22      5489916        5483728   -0.1%
   23      5503165        5526401   +0.4%
   24      5556469        5570199   +0.2%
   25      5525053        5510563   -0.3%
   26      5547215        5549489   +0.0%
   27      5484206        5486303   +0.0%
   28      5529157        5525227   -0.1%
   29      5525282        5538166   +0.2%
   30      5531291        5525207   -0.1%
   31      5546468        5543883   -0.0%
   32      5543200        5534168   -0.2%
   33      5509194        5507454   -0.0%
   34      5567784        5565927   -0.0%
   35      5566150        5562307   -0.1%
   36      5586535        5586133   -0.0%
   37      5515227        5514678   -0.0%
   38      5531903        5568247   +0.7%
   39      5465483        5508943   +0.8%
   40      5440502        5522920   +1.5%

             nodes    time  median nps  faster half
base       9312660  1.68 s     5528167      5547569
candidate  9312660  1.68 s     5528446      5550472
change                           +0.0%        +0.1%

paired change +0.0%, 95% interval -0.0% to +0.1%
diagnostic, on the default layout alone +0.1%, 95% interval -0.1% to +0.4%, 13 rounds

no change beyond ±1.0%: the whole interval is inside it

Speed: +0.0% (bench nps, 95% interval -0.0% to +0.1%, 40 interleaved rounds over shuffled layouts vs 3ab386ca)

The instructions each side's bench executed, counted under
cachegrind. The count repeats to within a few hundred
instructions, so a small change here is a real one, but it
prices instructions only: cache misses, mispredicted branches
and where the code lands are the speed's to show.

instructions against 3ab386ca, one cachegrind run a side, bench

             instructions    nodes  per node
base       13,159,580,971  9312660    1413.1
candidate  13,159,580,970  9312660    1413.1
change             -0.00%             -0.00%

within ±0.7%, as far as edits not made for speed have moved the count