cas: clang-tidy sweep, remaining checks (arm_tidy, T13 batch 2)

CI report: https://altinity-build-artifacts.s3.amazonaws.com/json.html?PR=2073&sha=aeb13b24394023fa8cd9d310d4cbcbc308380af1&name_0=PR&name_1=Build+(arm_tidy)
PR: #2073

Semantics-preserving conformance for the remaining flagged classes:
readability-container-contains, readability-isolate-declaration,
google-runtime-int (AWS SDK retry-API overrides keep `long` with targeted
NOLINT — the override contract owns the type), readability-duplicate-include,
cppcoreguidelines-init-variables, cert-msc, modernize-raw-string-literal,
modernize-use-starts-ends-with, bugprone-empty-catch (comments only — no new
behavior), googletest naming, bugprone-argument-comment,
bugprone-optional-value-conversion, bugprone-misplaced-widening-cast
(CasTypes.h site audited: not a real precision bug — the value is
range-validated to 0-5; cast made explicit without value change).
CasRefCowMap's own `contains` keeps its `find` with NOLINT (self-recursion).

Bulk edits by codex (gpt-5.6-luna) per the T13 brief
(.superpowers/sdd/task-13-batch2-report.md); one over-removed include
(PartFolderAccess.h) restored and verification by Claude. Battery 919/919.

Signed-off-by: Mikhail Filimonov <mfilimonov@altinity.com>

Author: filimonov

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/Backend/CasInMemoryBackend.cpp

@@ -131,7 +131,7 @@ HeadResult InMemoryBackend::head(const String & key)
 PutResult InMemoryBackend::putIfAbsent(const String & key, const String & bytes, const ObjectMeta & meta)
 {
     std::lock_guard lock(mutex_);
-    if (store_.count(key))
+    if (store_.contains(key))
         return {PutOutcome::PreconditionFailed, {}};
 
     Token t = mintToken();
@@ -276,7 +276,7 @@ PutResult InMemoryBackend::promoteStaged(const String & staging_key, const Strin
             "InMemoryBackend::promoteStaged: staging object {} is absent", staging_key);
 
     /// Write-once: a present destination is the "lost the race" signal, not an overwrite.
-    if (store_.count(blob_key))
+    if (store_.contains(blob_key))
         return {PutOutcome::PreconditionFailed, {}};
 
     /// Server-side copy: the destination bytes ARE the staging bytes; a fresh monotone token stands in
@@ -332,7 +332,7 @@ ListPage InMemoryBackend::list(const String & prefix, const String & cursor, siz
     size_t count = 0;
     while (it != store_.end() && count < limit)
     {
-        if (it->first.substr(0, prefix.size()) != prefix)
+        if (!it->first.starts_with(prefix))
             break;
 
         ListedKey lk;
@@ -345,7 +345,7 @@ ListPage InMemoryBackend::list(const String & prefix, const String & cursor, siz
     }
 
     // Set next_cursor if there are more keys in this prefix
-    if (!page.keys.empty() && it != store_.end() && it->first.substr(0, prefix.size()) == prefix)
+    if (!page.keys.empty() && it != store_.end() && it->first.starts_with(prefix))
         page.next_cursor = page.keys.back().key;
 
     return page;

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/Backend/CasObjectStorageBackend.cpp

@@ -45,13 +45,13 @@ namespace DB::Cas
 /// not this SDK loop. Every consultation is counted because it proves that the SDK considered the
 /// first attempt inconclusive or failed; otherwise the retry-consultation metric would remain zero
 /// even when the SDK reached this decision point.
-bool detail::SingleAttemptRetryStrategy::ShouldRetry(const Aws::Client::AWSError<Aws::Client::CoreErrors> &, long) const
+bool detail::SingleAttemptRetryStrategy::ShouldRetry(const Aws::Client::AWSError<Aws::Client::CoreErrors> &, long) const // NOLINT(google-runtime-int): AWS SDK virtual API requires `long`.
 {
     recordConditionalWriteSdkRetryConsidered();
     return false;
 }
 
-long detail::SingleAttemptRetryStrategy::CalculateDelayBeforeNextRetry(const Aws::Client::AWSError<Aws::Client::CoreErrors> &, long) const
+long detail::SingleAttemptRetryStrategy::CalculateDelayBeforeNextRetry(const Aws::Client::AWSError<Aws::Client::CoreErrors> &, long) const // NOLINT(google-runtime-int): AWS SDK virtual API requires `long`.
 {
     /// AWSClient prepares the delay before calling `ShouldRetry`, so this method is reached even
     /// though no retry will be made. Returning zero avoids needless backoff computation; it is never
@@ -1027,7 +1027,7 @@ ListPage ObjectStorageBackend::list(const String & prefix, const String & cursor
         all.reserve(children.size());
         for (const auto & child : children)
         {
-            if (child->relative_path.substr(0, physical_prefix.size()) != physical_prefix)
+            if (!child->relative_path.starts_with(physical_prefix))
                 continue;
             ListedKey lk;
             lk.key = child->relative_path.substr(strip.size());
@@ -1064,7 +1064,7 @@ ListPage ObjectStorageBackend::list(const String & prefix, const String & cursor
     for (; it->isValid(); it->next())
     {
         const auto child = it->current();
-        if (child->relative_path.substr(0, physical_prefix.size()) != physical_prefix)
+        if (!child->relative_path.starts_with(physical_prefix))
             continue;
 
         ListedKey lk;

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/Backend/CasObjectStorageBackend.h

@@ -40,10 +40,10 @@ class SingleAttemptRetryStrategy final : public Aws::Client::RetryStrategy
 {
 public:
     /// Record that the SDK considered retrying, then refuse the retry.
-    bool ShouldRetry(const Aws::Client::AWSError<Aws::Client::CoreErrors> &, long) const override;
+    bool ShouldRetry(const Aws::Client::AWSError<Aws::Client::CoreErrors> &, long) const override; // NOLINT(google-runtime-int): AWS SDK virtual API requires `long`.
     /// Return no delay; `ShouldRetry` rejects the attempt immediately afterward.
-    long CalculateDelayBeforeNextRetry(const Aws::Client::AWSError<Aws::Client::CoreErrors> &, long) const override;
-    long GetMaxAttempts() const override { return 1; }
+    long CalculateDelayBeforeNextRetry(const Aws::Client::AWSError<Aws::Client::CoreErrors> &, long) const override; // NOLINT(google-runtime-int): AWS SDK virtual API requires `long`.
+    long GetMaxAttempts() const override { return 1; } // NOLINT(google-runtime-int): AWS SDK virtual API requires `long`.
 };
 }
 #endif

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/Backend/CasRequestControl.cpp

@@ -330,7 +330,7 @@ CasWriteOutcome CasRequestController::putIfAbsentControlled(
 
         /// The committed incarnation's token, filled by whichever leg proves Committed below.
         Token committed_token;
-        CasWriteOutcome attempt_outcome;
+        CasWriteOutcome attempt_outcome{};
         try
         {
             const PutResult put = backend->putIfAbsent(key_s, bytes_s);

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/ContentAddressedMetadataStorage.cpp

@@ -225,7 +225,7 @@ void ContentAddressedMetadataStorage::runOneGcRoundForTest()
     /// scheduler per call would acquire the lease on the first call and then back off forever
     /// ("incumbent alive" - its own previous incarnation). Recreating the scheduler for every call
     /// would therefore make every round after the first a silent no-op.
-    Cas::CasGcScheduler * sched;
+    Cas::CasGcScheduler * sched = nullptr;
     {
         std::lock_guard lock(gc_scheduler_mutex);
         if (!gc_scheduler)
@@ -356,7 +356,7 @@ Cas::RoundReport ContentAddressedMetadataStorage::runGarbageCollectionRoundNow()
             "Garbage collection is not enabled on this content-addressed disk");
     /// Mirror runOneGcRoundForTest: a STABLE scheduler instance across calls (the lease's
     /// observation-window steal protocol compares consecutive observations of the same gc_id).
-    Cas::CasGcScheduler * sched;
+    Cas::CasGcScheduler * sched = nullptr;
     {
         std::lock_guard lock(gc_scheduler_mutex);
         if (!gc_scheduler)

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/ContentAddressedTransaction.cpp

@@ -246,7 +246,7 @@ void ContentAddressedTransaction::uploadPendingBlobs(PartStaging & st)
 
     for (const auto & pb : st.pending_blobs)         /// pool writes — uploads + 412/HEAD/resurrect
     {
-        if (!referenced_hashes.count(pb.ref))
+        if (!referenced_hashes.contains(pb.ref))
             continue;   /// The entry was removed by unlinkFile/replaceFile; skip this orphan.
         /// Each pending blob
         /// is promoted through the SAME condemn/resurrect gate in `PartWriteTxn::putBlob`. Only the upload

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/Formats/CasLayout.cpp

@@ -25,7 +25,7 @@ std::optional<BlobRef> Layout::parseBlobKey(std::string_view key) const
         rest.remove_suffix(kMetaSuffix.size());
 
     const String blobs_root = blobsPrefix();   /// "<prefix>/blobs/"
-    if (rest.size() <= blobs_root.size() || rest.substr(0, blobs_root.size()) != blobs_root)
+    if (rest.size() <= blobs_root.size() || !rest.starts_with(blobs_root))
         return std::nullopt;
     rest.remove_prefix(blobs_root.size());
 

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/Formats/CasLayout.h

@@ -166,7 +166,7 @@ class Layout
         if (ns_part.empty())
             return std::nullopt;
 
-        RefObjectKind kind;
+        RefObjectKind kind{};
         if (kind_seg == "_cleanup")
             kind = RefObjectKind::Cleanup;
         else if (kind_seg == "_log")

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/Formats/CasPartManifestFormat.cpp

@@ -139,7 +139,9 @@ PartManifest decodePartManifest(std::string_view data)
         const String meta = readLine(in, line_cap, "cas_part_manifest");
         ReadBufferFromMemory mm(meta.data(), meta.size());
         JsonObjectReader r(mm, KeyStrictness::Tolerant, "cas_part_manifest");
-        std::optional<uint64_t> me, mb, mo;
+        std::optional<uint64_t> me;
+        std::optional<uint64_t> mb;
+        std::optional<uint64_t> mo;
         std::optional<String> ns;
         std::optional<UInt128> pd;
         String key;
@@ -195,8 +197,11 @@ PartManifest decodePartManifest(std::string_view data)
             throw Exception(ErrorCodes::CORRUPTED_DATA, "PartManifest: record must start with \"p\"");
         ManifestEntry e;
         e.path = r.readString();
-        std::optional<String> pm, ha, h;
-        std::optional<uint64_t> sz, il;
+        std::optional<String> pm;
+        std::optional<String> ha;
+        std::optional<String> h;
+        std::optional<uint64_t> sz;
+        std::optional<uint64_t> il;
         while (r.nextKey(key))
         {
             if (key == "pm") pm = r.readString();

src/Disks/DiskObjectStorage/MetadataStorages/ContentAddressed/Formats/CasRecordStreamFormat.cpp

@@ -252,9 +252,17 @@ bool SourceEdgeRunReader::next(SourceEdgeRecord & rec)
     }
 
     SourceEdgeRecord out;
-    String b, tv;
-    bool have_b = false, have_s = false, have_m = false;
-    bool have_pend = false, have_tt = false, have_tv = false, have_sz = false, have_cr = false, have_mc = false;
+    String b;
+    String tv;
+    bool have_b = false;
+    bool have_s = false;
+    bool have_m = false;
+    bool have_pend = false;
+    bool have_tt = false;
+    bool have_tv = false;
+    bool have_sz = false;
+    bool have_cr = false;
+    bool have_mc = false;
     TokenType tt{};
     do
     {