fix: Podman compatibility and testing

.github/workflows/run_tests.yml

@@ -164,24 +164,37 @@ jobs:
 
     strategy:
       matrix:
-        os: [ubuntu-22.04]
+        os: [ubuntu-24.04]
         python-version: ["3.12"]
 
         arch: ["x64"]
+        docker-engine: ["docker"]
 
         unit-tesseract: ${{ fromJson(needs.get-e2e-matrix.outputs.matrix) }}
 
         include:
           # Test on arm to ensure compatibility with Apple M1 chips
           # (OSX runners don't have access to Docker so we use Linux ARM runners instead)
-          - os: "ubuntu-22.04"
+          - os: "ubuntu-24.04"
             python-version: "3.12"
             arch: "arm"
+            docker-engine: "docker"
             unit-tesseract: "base"
+          # Run tests using Podman
+          - os: "ubuntu-24.04"
+            python-version: "3.12"
+            arch: "x64"
+            docker-engine: "podman"
+            unit-tesseract: "base"
+          - os: "ubuntu-24.04"
+            python-version: "3.12"
+            arch: "x64"
+            docker-engine: "podman"
+            unit-tesseract: "vectoradd_jax"
 
       fail-fast: false
 
-    runs-on: ${{ matrix.arch == 'x64' && matrix.os == 'ubuntu-22.04' && 'ubuntu-22.04' || matrix.arch == 'arm' && matrix.os == 'ubuntu-22.04' && 'linux-arm64-ubuntu2204' }}
+    runs-on: ${{ matrix.arch == 'x64' && matrix.os == 'ubuntu-24.04' && 'ubuntu-24.04' || matrix.arch == 'arm' && matrix.os == 'ubuntu-24.04' && 'linux-arm64-ubuntu2404' ||  matrix.arch == 'x64' && matrix.os == 'ubuntu-24.04' && 'ubuntu-24.04'}}
 
     defaults:
       run:
@@ -209,8 +222,27 @@ jobs:
         run: |
           uv sync --extra dev --frozen
 
+      - name: Set up Podman
+        if: matrix.docker-engine == 'podman'
+        run: |
+          systemctl start --user podman.socket
+          systemctl status --user podman.socket
+
+          podman --version
+
+          export DOCKER_HOST=unix:///run/user/1001/podman/podman.sock
+          echo "DOCKER_HOST=${DOCKER_HOST}" >> $GITHUB_ENV
+
+      - name: Free some disk space
+        run: |
+          sudo rm -rf /usr/share/dotnet /usr/local/lib/android /opt/ghc \
+                      /usr/local/share/powershell \
+                      /usr/share/swift /usr/local/.ghcup /usr/lib/jvm
+
       - name: Run test suite
         run: |
+          env | grep DOCKER_HOST || true
+
           if [ "${{ matrix.unit-tesseract }}" == "base" ]; then
             uv run --no-sync pytest \
               --always-run-endtoend \

tesseract_core/sdk/docker_client.py

@@ -42,10 +42,48 @@ def from_dict(cls, json_dict: dict) -> "Image":
 class Images:
     """Namespace for functions to interface with Tesseract docker images."""
 
+    @staticmethod
+    def _tag_exists(image_name: str, tags: list_) -> bool:
+        """Helper function to check if image name exists in the list of tags.
+
+        Specially handling has to be done to achieve unfuzzy substring matching, i.e.
+        if image tag is foo/bar/image, we need to return true for foo/bar/image, bar/image,
+        and image, but not ar/image.
+
+        There is no equivalent in docker-py.
+
+        Params:
+            image_name: The image name to check.
+            tags: The list of tags to check against.
+
+        Returns:
+            True if the image name exists in the list of tags, False otherwise.
+        """
+        if ":" not in image_name:
+            image_name += ":latest"
+
+        image_name_parts = image_name.strip("/").split("/")
+        for tag in tags:
+            tag_parts = tag.strip("/").split("/")
+            # Check if the last parts of the tag matches the subportion that image name specifies
+            if tag_parts[-len(image_name_parts) :] == image_name_parts:
+                return True
+        return False
+
     @staticmethod
     def get(image_id_or_name: str | bytes, tesseract_only: bool = True) -> Image:
         """Returns the metadata for a specific image.
 
+        In docker-py, there is no substring matching and the image name is the
+        last tag in the list of tags, so if an image has multiple tags, only
+        one of the tags would be able to find the image.
+
+        However, in podman, this is not the case. Podman has substring matching
+        by "/" segments to handle repository urls and returns images even if
+        partial name is specified, or if image has multiple tags.
+
+        We chose to support podman's largest string matching functionality here.
+
         Params:
             image_id_or_name: The image name or id to get.
             tesseract_only: If True, only retrieves Tesseract images.
@@ -77,11 +115,7 @@ def is_image_id(s: str) -> bool:
                 image_obj.id == image_id_or_name
                 or image_obj.short_id == image_id_or_name
                 or image_id_or_name in image_obj.tags
-                or (
-                    any(
-                        tag.split("/")[-1] == image_id_or_name for tag in image_obj.tags
-                    )
-                )
+                or Images._tag_exists(image_id_or_name, image_obj.tags)
             ):
                 return image_obj
 
@@ -711,7 +745,7 @@ class DockerException(Exception):
 class BuildError(DockerException):
     """Raised when a build fails."""
 
-    def __init__(self, build_log: str) -> None:
+    def __init__(self, build_log: list_[str]) -> None:
         self.build_log = build_log
 
 

tests/endtoend_tests/common.py

@@ -48,7 +48,7 @@ def print_debug_info(result):
 
 
 def build_tesseract(
-    sourcedir, image_name, config_override=None, tag=None, build_retries=3
+    client, sourcedir, image_name, config_override=None, tag=None, build_retries=3
 ):
     cli_runner = CliRunner(mix_stderr=False)
 
@@ -87,6 +87,7 @@ def build_tesseract(
     assert result.exit_code == 0, result.exception
 
     image_tags = json.loads(result.stdout.strip())
-    assert image_name in image_tags
+
+    assert client.images._tag_exists(image_name, image_tags)
 
     return image_tags[0]

tests/endtoend_tests/test_docker_client.py

@@ -3,6 +3,7 @@
 
 """End to end tests for docker cli wrapper."""
 
+import os
 import subprocess
 import textwrap
 from contextlib import closing
@@ -52,7 +53,7 @@ def test_get_image(docker_client, docker_client_built_image_name, docker_py_clie
     # Get the image
     image = docker_client.images.get(docker_client_built_image_name)
     assert image is not None
-    assert docker_client_built_image_name in image.tags
+    assert docker_client.images._tag_exists(docker_client_built_image_name, image.tags)
 
     docker_py_image = docker_py_client.images.get(docker_client_built_image_name)
     assert docker_py_image is not None
@@ -100,7 +101,7 @@ def test_get_image(docker_client, docker_client_built_image_name, docker_py_clie
     assert docker_py_image_list is not None
     # Check that every image in image_list is also in docker_py_image_list
     for image in image_list:
-        assert image.id in [img.id for img in docker_py_image_list]
+        assert image.id in [img.id for img in docker_py_image_list if img]
 
 
 def test_create_image(
@@ -120,7 +121,9 @@ def test_create_image(
     try:
         image = docker_client.images.get(docker_client_built_image_name)
         assert image is not None
-        assert docker_client_built_image_name in image.tags
+        assert docker_client.images._tag_exists(
+            docker_client_built_image_name, image.tags
+        )
 
         image_id_obj = docker_client.images.get(image.id)
         image_short_id_obj = docker_client.images.get(image.short_id)
@@ -138,7 +141,7 @@ def test_create_image(
         image1 = docker_client.images.get(image1_name)
         image1_name = image1_name + ":latest"
         assert image1 is not None
-        assert image1_name in image1.tags
+        assert docker_client.images._tag_exists(image1_name, image1.tags)
         assert image_exists(docker_client, image1_name)
         assert image_exists(docker_py_client, image1_name)
 
@@ -155,14 +158,33 @@ def test_create_image(
         # Our docker client should be able to retrieve images with just the name
         image2 = docker_client.images.get(repo_url + image2_name)
         image2_no_url = docker_client.images.get(image2_name)
+        assert docker_client.images._tag_exists(image2_name, image2.tags)
+
         assert image2 is not None
         assert image2_no_url is not None
         assert image2.id == image2_py.id
         assert image2_no_url.id == image2_py.id
 
-        assert image_exists(docker_client, image2_name)
-        assert image_exists(docker_client, repo_url + image2_name)
-        assert image_exists(docker_py_client, repo_url + image2_name)
+        # Check we are not overmatching but we are getting all possible cases
+        docker_host = os.environ.get("DOCKER_HOST", "")
+
+        podman = False
+        if "podman" in docker_host:
+            podman = True
+
+        for client in [docker_client, docker_py_client]:
+            assert not image_exists(client, "create_image")
+
+            if podman and client == docker_py_client:
+                # Docker-py does not support partial string matching
+                assert image_exists(client, image2_name)
+                assert image_exists(client, f"/{image2_name}")
+                assert image_exists(client, f"bar/{image2_name}")
+                assert image_exists(client, f"bar/{image2_name}:latest")
+                assert not image_exists(client, f"ar/{image2_name}")
+                assert image_exists(client, f"foo/bar/{image2_name}")
+
+            assert image_exists(client, repo_url + image2_name)
 
     finally:
         # Clean up the images

tests/endtoend_tests/test_endtoend.py

@@ -4,6 +4,7 @@
 """End-to-end tests for Tesseract workflows."""
 
 import json
+import os
 from pathlib import Path
 
 import pytest
@@ -20,7 +21,9 @@ def built_image_name(
     docker_client, docker_cleanup, shared_dummy_image_name, dummy_tesseract_location
 ):
     """Build the dummy Tesseract image for the tests."""
-    image_name = build_tesseract(dummy_tesseract_location, shared_dummy_image_name)
+    image_name = build_tesseract(
+        docker_client, dummy_tesseract_location, shared_dummy_image_name
+    )
     assert image_exists(docker_client, image_name)
     docker_cleanup["images"].append(image_name)
     yield image_name
@@ -59,7 +62,11 @@ def test_build_from_init_endtoend(
         config_override["build_config.base_image"] = base_image
 
     image_name = build_tesseract(
-        tmp_path, dummy_image_name, config_override=config_override, tag=img_tag
+        docker_client,
+        tmp_path,
+        dummy_image_name,
+        config_override=config_override,
+        tag=img_tag,
     )
     assert image_exists(docker_client, image_name)
     docker_cleanup["images"].append(image_name)
@@ -318,6 +325,14 @@ def test_tesseract_serve_ports(built_image_name, port, docker_cleanup):
     cli_runner = CliRunner(mix_stderr=False)
     project_id = None
 
+    docker_host = os.environ.get("DOCKER_HOST", "")
+
+    if "-" in port and "podman" in docker_host:
+        pytest.skip(
+            "Podman does not support port ranges in compose."
+            "See https://github.com/containers/podman/issues/15111"
+        )
+
     # Serve tesseract on specified ports.
     run_res = cli_runner.invoke(
         app,

tests/endtoend_tests/test_examples.py

@@ -21,7 +21,7 @@
 import numpy.typing as npt
 import pytest
 import requests
-from common import build_tesseract
+from common import build_tesseract, image_exists
 from typer.testing import CliRunner
 
 
@@ -739,15 +739,6 @@ def unit_tesseract_config(unit_tesseract_names, unit_tesseract_path):
     return TEST_CASES[unit_tesseract_path.name]
 
 
-def image_exists(client, image_name):
-    # Docker images may be prefixed with the registry URL
-    return any(
-        tag.split("/")[-1] == image_name
-        for img in client.images.list()
-        for tag in img.tags
-    )
-
-
 def print_debug_info(result):
     """Print debug info from result of a CLI command if it failed."""
     if result.exit_code == 0:
@@ -816,6 +807,7 @@ def test_unit_tesseract_endtoend(
 
     # Stage 1: Build
     img_name = build_tesseract(
+        docker_client,
         unit_tesseract_path,
         dummy_image_name,
         tag="sometag",

tests/endtoend_tests/test_tesseract_sdk.py

@@ -27,7 +27,9 @@ def built_image_name(
     docker_client, shared_dummy_image_name, dummy_tesseract_location, docker_cleanup
 ):
     """Build the dummy Tesseract image for the tests."""
-    image_name = build_tesseract(dummy_tesseract_location, shared_dummy_image_name)
+    image_name = build_tesseract(
+        docker_client, dummy_tesseract_location, shared_dummy_image_name
+    )
     assert image_exists(docker_client, image_name)
     docker_cleanup["images"].append(image_name)
     yield image_name