ci: handle large repo mirroring via GitLab import

Add size-aware mirroring logic that routes large empty projects through GitLab’s server-side import API before pushing, to avoid GitLab.com’s 5 GiB push limit. The workflow now tracks project emptiness, starts authenticated imports for private sources, waits for import completion with clear status/permission/time-out handling, and then proceeds with mirror push. It also removes GitHub-only `refs/pull/*` refs from the local mirror before push.

Author: ReenigneArcher

.github/workflows/__mirror-github-to-gitlab.yml

@@ -56,6 +56,7 @@ jobs:
             const repositoryData = repos.map((repo) => ({
               cloneUrl: repo.clone_url,
               name: repo.name,
+              sizeKiB: repo.size,
               targetVisibility: repo.visibility === 'public' ? 'public' : 'private',
               ...gitlabTarget(repo),
             }));
@@ -149,6 +150,57 @@ jobs:
             curl "${curl_args[@]}" "${url}"
           }
 
+          wait_for_gitlab_import() {
+            local project_id="$1"
+            local source_name="$2"
+            local deadline="$((SECONDS + 10800))"
+            local import_error
+            local import_status
+            local status
+
+            while ((SECONDS < deadline)); do
+              if ! status="$(gitlab_request GET "${GITLAB_API_URL}/projects/${project_id}/import")"; then
+                echo "::error title=Mirror failed: ${source_name}::Unable to query the GitLab import."
+                return 1
+              fi
+              if [[ "${status}" != "200" ]]; then
+                if [[ "${status}" == "403" ]]; then
+                  echo "::error title=Mirror failed: ${source_name}::GITLAB_TOKEN needs Import API Read access."
+                  return 1
+                fi
+                echo "::error title=Mirror failed: ${source_name}::Unable to query the GitLab import (HTTP ${status})."
+                return 1
+              fi
+
+              import_status="$(jq -er '.import_status' "${response_file}")"
+              case "${import_status}" in
+                finished)
+                  echo "GitLab server-side import finished."
+                  return 0
+                  ;;
+                scheduled | started)
+                  echo "GitLab server-side import status: ${import_status}."
+                  sleep 30
+                  ;;
+                failed)
+                  import_error="$(jq -r '.import_error // "No import error detail was provided."' "${response_file}")"
+                  import_error="${import_error//$'\r'/ }"
+                  import_error="${import_error//$'\n'/ }"
+                  echo "GitLab import failed: ${import_error}"
+                  echo "::error title=Mirror failed: ${source_name}::GitLab server-side import failed."
+                  return 1
+                  ;;
+                *)
+                  echo "::error title=Mirror failed: ${source_name}::Unexpected GitLab import status: ${import_status}."
+                  return 1
+                  ;;
+              esac
+            done
+
+            echo "::error title=Mirror failed: ${source_name}::GitLab server-side import timed out after three hours."
+            return 1
+          }
+
           encoded_group="$(jq -rn --arg value "${GITLAB_GROUP}" '$value | @uri')"
           if ! status="$(gitlab_request GET "${GITLAB_API_URL}/groups/${encoded_group}")"; then
             echo "::error::Unable to query the GitLab group."
@@ -163,12 +215,17 @@ jobs:
           github_auth="$(printf 'x-access-token:%s' "${GITHUB_TOKEN}" | base64 --wrap=0)"
           echo "::add-mask::${github_auth}"
 
+          # GitLab.com limits each Git push to 5 GiB. Seed large empty projects through GitLab's import API,
+          # using a lower threshold because GitHub's reported repository size is not the resulting pack size.
+          gitlab_import_threshold_kib="$((4 * 1024 * 1024))"
+
           repository_count="$(jq -er 'length' "${repository_file}")"
           for ((index = 0; index < repository_count; index++)); do
             repository="$(jq -ec ".[${index}]" "${repository_file}")"
             repository_number="$((index + 1))"
             source_name="$(jq -er '.name' <<< "${repository}")"
             source_clone_url="$(jq -er '.cloneUrl' <<< "${repository}")"
+            source_size_kib="$(jq -er '.sizeKiB' <<< "${repository}")"
             target_name="$(jq -er '.targetName' <<< "${repository}")"
             target_path="$(jq -er '.targetPath' <<< "${repository}")"
             target_visibility="$(jq -er '.targetVisibility' <<< "${repository}")"
@@ -180,6 +237,8 @@ jobs:
             echo "Visibility: ${target_visibility}"
             echo "Checking for the GitLab project..."
 
+            project_created=false
+
             if ! status="$(gitlab_request GET "${GITLAB_API_URL}/projects/${encoded_project_path}")"; then
               echo "::error title=Mirror failed: ${source_name}::Unable to query the GitLab project."
               exit 1
@@ -210,6 +269,7 @@ jobs:
                 echo "::error title=Mirror failed: ${source_name}::${error_message}"
                 exit 1
               fi
+              project_created=true
             elif [[ "${status}" != "200" ]]; then
               echo "::error title=Mirror failed: ${source_name}::Unable to query the GitLab project (HTTP ${status})."
               exit 1
@@ -218,6 +278,13 @@ jobs:
             fi
 
             project_id="$(jq -er '.id' "${response_file}")"
+            if [[ "${project_created}" == "true" ]]; then
+              project_empty=true
+            else
+              project_empty="$(
+                jq -r '(.empty_repo == true) or (.default_branch == null)' "${response_file}"
+              )"
+            fi
 
             # Make a private source private before changing metadata or pushing any Git data.
             if [[ "${target_visibility}" == "private" ]]; then
@@ -267,6 +334,52 @@ jobs:
 
             target_clone_url="$(jq -er '.ssh_url_to_repo' "${response_file}")"
 
+            gitlab_import_started=false
+            if [[ "${project_empty}" == "true" ]] && \
+              ((source_size_kib >= gitlab_import_threshold_kib)); then
+              echo "Empty large project detected (${source_size_kib} KiB reported by GitHub)."
+              echo "Scheduling a GitLab server-side Git import to avoid the 5 GiB push limit..."
+              if [[ "${target_visibility}" == "private" ]]; then
+                import_payload="$(
+                  jq -nc \
+                    --arg import_url "${source_clone_url}" \
+                    --arg import_url_user "x-access-token" \
+                    --arg import_url_password "${GITHUB_TOKEN}" \
+                    '{
+                      import_url: $import_url,
+                      import_url_user: $import_url_user,
+                      import_url_password: $import_url_password
+                    }'
+                )"
+              else
+                import_payload="$(
+                  jq -nc \
+                    --arg import_url "${source_clone_url}" \
+                    '{import_url: $import_url}'
+                )"
+              fi
+
+              if ! status="$(
+                gitlab_request POST "${GITLAB_API_URL}/projects/${project_id}/import/git" "${import_payload}"
+              )"; then
+                echo "::error title=Mirror failed: ${source_name}::Unable to schedule the GitLab import."
+                exit 1
+              fi
+              if [[ "${status}" == "403" ]]; then
+                echo "::error title=Mirror failed: ${source_name}::GITLAB_TOKEN needs Import API Create access."
+                exit 1
+              fi
+              if [[ "${status}" != "201" ]] && [[ "${status}" != "202" ]]; then
+                echo "::error title=Mirror failed: ${source_name}::Unable to schedule the GitLab import" \
+                  "(HTTP ${status})."
+                exit 1
+              fi
+              gitlab_import_started=true
+            elif [[ "${project_empty}" != "true" ]] && \
+              ((source_size_kib >= gitlab_import_threshold_kib)); then
+              echo "Large project is already populated; using an incremental Git push."
+            fi
+
             mirror_dir="${temp_root}/repository.git"
             echo "Cloning ${source_clone_url} as a mirror..."
             if ! git \
@@ -275,6 +388,22 @@ jobs:
               echo "::error title=Mirror failed: ${source_name}::Unable to clone the GitHub repository."
               exit 1
             fi
+
+            mapfile -t pull_refs < <(
+              git -C "${mirror_dir}" for-each-ref --format='%(refname)' refs/pull/
+            )
+            if ((${#pull_refs[@]} > 0)); then
+              echo "Removing ${#pull_refs[@]} GitHub-only pull request refs from the mirror..."
+              printf 'delete %s\n' "${pull_refs[@]}" | git -C "${mirror_dir}" update-ref --stdin
+            fi
+
+            if [[ "${gitlab_import_started}" == "true" ]]; then
+              echo "Waiting for the GitLab server-side import to finish..."
+              if ! wait_for_gitlab_import "${project_id}" "${source_name}"; then
+                exit 1
+              fi
+            fi
+
             echo "Pushing the mirror to ${target_clone_url}..."
             if ! git \
               -C "${mirror_dir}" \