Set FSLOUTPUTTYPE in container env, drop styxcache bypass workaround

The nightly surfaced a real latent bug once we got container stderr
flowing: lta_convert's FSL output path exits non-zero with "ERROR:
trying to convert FreeSurfer registration matrix to FSL, but
FSLOUTPUTTYPE variable is not set." The default env dict only injected
ITK threads, ANTs seed, and SURFER_SIDEDOOR; nothing set the FSL
output type, which FreeSurfer (and FSL) reads to decide what
filesystem format to emit. Adding FSLOUTPUTTYPE=NIFTI_GZ is idiomatic
FSL default behavior and is ignored by non-FSL tools, so applying it
globally is safe.

Also:
- Bump styx-runtime-py packages: styxpodman 0.1.1->0.2.0 (carries the
  fix from styx-api/styx-runtime-py#9 that surfaces container stderr
  through the cache wrapper on failure), plus siblings.
- Drop the bypass_cache kwarg and RBC_STYXCACHE_DIR clearing from the
  template fixture. With the new styxpodman, the raw tool stderr now
  reaches the parent subprocess even through the cache wrapper, so
  the workaround is obsolete and we can keep the cache speedup on
  this invocation too.

Author: nx10

src/rbc/orchestration/__init__.py

@@ -24,6 +24,11 @@
 _DEFAULT_ENV_VARS = {
     "ITK_GLOBAL_DEFAULT_NUMBER_OF_THREADS": "1",
     "ANTS_RANDOM_SEED": CPAC_ANTS_SEED,
+    # Required by any FreeSurfer tool that emits FSL-format output
+    # (e.g. lta_convert --outfsl): without it FS exits non-zero with
+    # "FSLOUTPUTTYPE variable is not set". Harmless for tools that
+    # ignore it. NIFTI_GZ matches FSL's conventional default.
+    "FSLOUTPUTTYPE": "NIFTI_GZ",
 }
 
 

tests/integration/longitudinal/conftest.py

@@ -7,7 +7,6 @@
 
 from __future__ import annotations
 
-import os
 import shutil
 import subprocess
 from pathlib import Path
@@ -35,30 +34,19 @@ def _rbc_exe() -> str:
 
 
 def _run_rbc(
-    args: Sequence[str], *, timeout: int = 7200, bypass_cache: bool = False
+    args: Sequence[str], *, timeout: int = 7200
 ) -> subprocess.CompletedProcess[str]:
     """Invoke ``rbc`` with ``-vv`` so container output reaches the subprocess.
 
     Without ``-vv`` the styx runner logger stays at WARNING and
     container-side output never makes it into the captured stderr, which
     leaves the test with only a bare ``returncode=1`` to diagnose from.
-
-    Set ``bypass_cache=True`` to run with ``RBC_STYXCACHE_DIR`` cleared.
-    styxcache's tee handlers swallow container stderr into an internal
-    gzip that's discarded on exception, so a failing niwrap call under
-    cache leaves no diagnostic trail; bypassing cache restores raw
-    podman stderr passthrough.
     """
-    env = os.environ.copy()
-    if bypass_cache:
-        env.pop("RBC_STYXCACHE_DIR", None)
-
     result = subprocess.run(  # noqa: S603
         [_rbc_exe(), *args, "-vv"],
         capture_output=True,
         text=True,
         timeout=timeout,
-        env=env,
     )
     assert result.returncode == 0, (
         f"rbc {args[0]} exited with code {result.returncode}\n"
@@ -158,11 +146,5 @@ def longitudinal_template_output(
             "--participant-label",
             _SUB,
         ],
-        # Template generation is still returning silent non-zero on the
-        # nightly; cache-bypass here gives us raw container stderr on
-        # failure without losing the anat fixture's cache benefit.
-        # TODO: drop once styx-api/styx-runtime-py#9 lands and surfaces
-        # container stderr through the cache wrapper.
-        bypass_cache=True,
     )
     return ds000114_anat_derivatives