Merge remote-tracking branch 'upstream/master' into no-stringly-typed-derivation-output

src/libfetchers/git.cc

@@ -282,7 +282,10 @@ struct GitInput : Input
                 // FIXME: git stderr messes up our progress indicator, so
                 // we're using --quiet for now. Should process its stderr.
                 try {
-                    runProgram("git", true, { "-C", repoDir, "fetch", "--quiet", "--force", "--", actualUrl, fmt("%s:%s", *input->ref, *input->ref) });
+                    auto fetchRef = input->ref->compare(0, 5, "refs/") == 0
+                        ? *input->ref
+                        : "refs/heads/" + *input->ref;
+                    runProgram("git", true, { "-C", repoDir, "fetch", "--quiet", "--force", "--", actualUrl, fmt("%s:%s", fetchRef, fetchRef) });
                 } catch (Error & e) {
                     if (!pathExists(localRefFile)) throw;
                     warn("could not update local clone of Git repository '%s'; continuing with the most recent version", actualUrl);
@@ -418,7 +421,7 @@ struct GitInputScheme : InputScheme
 
         auto input = std::make_unique<GitInput>(parseURL(getStrAttr(attrs, "url")));
         if (auto ref = maybeGetStrAttr(attrs, "ref")) {
-            if (!std::regex_match(*ref, refRegex))
+            if (std::regex_search(*ref, badGitRefRegex))
                 throw BadURL("invalid Git branch/tag name '%s'", *ref);
             input->ref = *ref;
         }

src/libfetchers/tarball.cc

@@ -71,7 +71,8 @@ DownloadFileResult downloadFile(
         info.narHash = hashString(htSHA256, *sink.s);
         info.narSize = sink.s->size();
         info.ca = makeFixedOutputCA(FileIngestionMethod::Flat, hash);
-        store->addToStore(info, sink.s, NoRepair, NoCheckSigs);
+        auto source = StringSource { *sink.s };
+        store->addToStore(info, source, NoRepair, NoCheckSigs);
         storePath = std::move(info.path);
     }
 

src/libstore/binary-cache-store.cc

@@ -113,9 +113,12 @@ void BinaryCacheStore::writeNarInfo(ref<NarInfo> narInfo)
         diskCache->upsertNarInfo(getUri(), hashPart, std::shared_ptr<NarInfo>(narInfo));
 }
 
-void BinaryCacheStore::addToStore(const ValidPathInfo & info, const ref<std::string> & nar,
+void BinaryCacheStore::addToStore(const ValidPathInfo & info, Source & narSource,
     RepairFlag repair, CheckSigsFlag checkSigs, std::shared_ptr<FSAccessor> accessor)
 {
+    // FIXME: See if we can use the original source to reduce memory usage.
+    auto nar = make_ref<std::string>(narSource.drain());
+
     if (!repair && isValidPath(info.path)) return;
 
     /* Verify that all references are valid. This may do some .narinfo
@@ -347,7 +350,8 @@ StorePath BinaryCacheStore::addToStore(const string & name, const Path & srcPath
 
     ValidPathInfo info(makeFixedOutputPath(method, h, name));
 
-    addToStore(info, sink.s, repair, CheckSigs, nullptr);
+    auto source = StringSource { *sink.s };
+    addToStore(info, source, repair, CheckSigs, nullptr);
 
     return std::move(info.path);
 }
@@ -361,7 +365,8 @@ StorePath BinaryCacheStore::addTextToStore(const string & name, const string & s
     if (repair || !isValidPath(info.path)) {
         StringSink sink;
         dumpString(s, sink);
-        addToStore(info, sink.s, repair, CheckSigs, nullptr);
+        auto source = StringSource { *sink.s };
+        addToStore(info, source, repair, CheckSigs, nullptr);
     }
 
     return std::move(info.path);

src/libstore/binary-cache-store.hh

@@ -74,7 +74,7 @@ public:
     std::optional<StorePath> queryPathFromHashPart(const std::string & hashPart) override
     { unsupported("queryPathFromHashPart"); }
 
-    void addToStore(const ValidPathInfo & info, const ref<std::string> & nar,
+    void addToStore(const ValidPathInfo & info, Source & narSource,
         RepairFlag repair, CheckSigsFlag checkSigs,
         std::shared_ptr<FSAccessor> accessor) override;
 

src/libstore/export-import.cc

@@ -1,3 +1,4 @@
+#include "serialise.hh"
 #include "store-api.hh"
 #include "archive.hh"
 #include "worker-protocol.hh"
@@ -100,7 +101,9 @@ StorePaths Store::importPaths(Source & source, std::shared_ptr<FSAccessor> acces
         if (readInt(source) == 1)
             readString(source);
 
-        addToStore(info, tee.source.data, NoRepair, checkSigs, accessor);
+        // Can't use underlying source, which would have been exhausted
+        auto source = StringSource { *tee.source.data };
+        addToStore(info, source, NoRepair, checkSigs, accessor);
 
         res.push_back(info.path.clone());
     }

src/libstore/store-api.cc

@@ -833,21 +833,6 @@ std::string makeFixedOutputCA(FileIngestionMethod method, const Hash & hash)
 }
 
 
-void Store::addToStore(const ValidPathInfo & info, Source & narSource,
-    RepairFlag repair, CheckSigsFlag checkSigs,
-    std::shared_ptr<FSAccessor> accessor)
-{
-    addToStore(info, make_ref<std::string>(narSource.drain()), repair, checkSigs, accessor);
-}
-
-void Store::addToStore(const ValidPathInfo & info, const ref<std::string> & nar,
-    RepairFlag repair, CheckSigsFlag checkSigs,
-    std::shared_ptr<FSAccessor> accessor)
-{
-    StringSource source(*nar);
-    addToStore(info, source, repair, checkSigs, accessor);
-}
-
 }
 
 

src/libstore/store-api.hh

@@ -451,12 +451,7 @@ public:
     /* Import a path into the store. */
     virtual void addToStore(const ValidPathInfo & info, Source & narSource,
         RepairFlag repair = NoRepair, CheckSigsFlag checkSigs = CheckSigs,
-        std::shared_ptr<FSAccessor> accessor = 0);
+        std::shared_ptr<FSAccessor> accessor = 0) = 0;
-
-    // FIXME: remove
-    virtual void addToStore(const ValidPathInfo & info, const ref<std::string> & nar,
-        RepairFlag repair = NoRepair, CheckSigsFlag checkSigs = CheckSigs,
-        std::shared_ptr<FSAccessor> accessor = 0);
 
     /* Copy the contents of a path to the store and register the
        validity the resulting path.  The resulting path is returned.

src/libutil/args.cc

@@ -217,10 +217,15 @@ MultiCommand::MultiCommand(const Commands & commands)
 {
     expectedArgs.push_back(ExpectedArg{"command", 1, true, [=](std::vector<std::string> ss) {
         assert(!command);
-        auto i = commands.find(ss[0]);
+        auto cmd = ss[0];
+        if (auto alias = get(deprecatedAliases, cmd)) {
+            warn("'%s' is a deprecated alias for '%s'", cmd, *alias);
+            cmd = *alias;
+        }
+        auto i = commands.find(cmd);
         if (i == commands.end())
-            throw UsageError("'%s' is not a recognised command", ss[0]);
+            throw UsageError("'%s' is not a recognised command", cmd);
-        command = {ss[0], i->second()};
+        command = {cmd, i->second()};
     }});
 
     categories[Command::catDefault] = "Available commands";

src/libutil/args.hh

@@ -234,6 +234,8 @@ public:
 
     std::map<Command::Category, std::string> categories;
 
+    std::map<std::string, std::string> deprecatedAliases;
+
     // Selected command, if any.
     std::optional<std::pair<std::string, ref<Command>>> command;
 

src/libutil/lru-cache.hh

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <cassert>
 #include <map>
 #include <list>
 #include <optional>

src/libutil/tests/lru-cache.cc

@@ -0,0 +1,130 @@
+#include "lru-cache.hh"
+#include <gtest/gtest.h>
+
+namespace nix {
+
+    /* ----------------------------------------------------------------------------
+     * size
+     * --------------------------------------------------------------------------*/
+
+    TEST(LRUCache, sizeOfEmptyCacheIsZero) {
+        LRUCache<std::string, std::string> c(10);
+        ASSERT_EQ(c.size(), 0);
+    }
+
+    TEST(LRUCache, sizeOfSingleElementCacheIsOne) {
+        LRUCache<std::string, std::string> c(10);
+        c.upsert("foo", "bar");
+        ASSERT_EQ(c.size(), 1);
+    }
+
+    /* ----------------------------------------------------------------------------
+     * upsert / get
+     * --------------------------------------------------------------------------*/
+
+    TEST(LRUCache, getFromEmptyCache) {
+        LRUCache<std::string, std::string> c(10);
+        auto val = c.get("x");
+        ASSERT_EQ(val.has_value(), false);
+    }
+
+    TEST(LRUCache, getExistingValue) {
+        LRUCache<std::string, std::string> c(10);
+        c.upsert("foo", "bar");
+        auto val = c.get("foo");
+        ASSERT_EQ(val, "bar");
+    }
+
+    TEST(LRUCache, getNonExistingValueFromNonEmptyCache) {
+        LRUCache<std::string, std::string> c(10);
+        c.upsert("foo", "bar");
+        auto val = c.get("another");
+        ASSERT_EQ(val.has_value(), false);
+    }
+
+    TEST(LRUCache, upsertOnZeroCapacityCache) {
+        LRUCache<std::string, std::string> c(0);
+        c.upsert("foo", "bar");
+        auto val = c.get("foo");
+        ASSERT_EQ(val.has_value(), false);
+    }
+
+    TEST(LRUCache, updateExistingValue) {
+        LRUCache<std::string, std::string> c(1);
+        c.upsert("foo", "bar");
+
+        auto val = c.get("foo");
+        ASSERT_EQ(val.value_or("error"), "bar");
+        ASSERT_EQ(c.size(), 1);
+
+        c.upsert("foo", "changed");
+        val = c.get("foo");
+        ASSERT_EQ(val.value_or("error"), "changed");
+        ASSERT_EQ(c.size(), 1);
+    }
+
+    TEST(LRUCache, overwriteOldestWhenCapacityIsReached) {
+        LRUCache<std::string, std::string> c(3);
+        c.upsert("one", "eins");
+        c.upsert("two", "zwei");
+        c.upsert("three", "drei");
+
+        ASSERT_EQ(c.size(), 3);
+        ASSERT_EQ(c.get("one").value_or("error"), "eins");
+
+        // exceed capacity
+        c.upsert("another", "whatever");
+
+        ASSERT_EQ(c.size(), 3);
+        // Retrieving "one" makes it the most recent element thus
+        // two will be the oldest one and thus replaced.
+        ASSERT_EQ(c.get("two").has_value(), false);
+        ASSERT_EQ(c.get("another").value(), "whatever");
+    }
+
+    /* ----------------------------------------------------------------------------
+     * clear
+     * --------------------------------------------------------------------------*/
+
+    TEST(LRUCache, clearEmptyCache) {
+        LRUCache<std::string, std::string> c(10);
+        c.clear();
+        ASSERT_EQ(c.size(), 0);
+    }
+
+    TEST(LRUCache, clearNonEmptyCache) {
+        LRUCache<std::string, std::string> c(10);
+        c.upsert("one", "eins");
+        c.upsert("two", "zwei");
+        c.upsert("three", "drei");
+        ASSERT_EQ(c.size(), 3);
+        c.clear();
+        ASSERT_EQ(c.size(), 0);
+    }
+
+    /* ----------------------------------------------------------------------------
+     * erase
+     * --------------------------------------------------------------------------*/
+
+    TEST(LRUCache, eraseFromEmptyCache) {
+        LRUCache<std::string, std::string> c(10);
+        ASSERT_EQ(c.erase("foo"), false);
+        ASSERT_EQ(c.size(), 0);
+    }
+
+    TEST(LRUCache, eraseMissingFromNonEmptyCache) {
+        LRUCache<std::string, std::string> c(10);
+        c.upsert("one", "eins");
+        ASSERT_EQ(c.erase("foo"), false);
+        ASSERT_EQ(c.size(), 1);
+        ASSERT_EQ(c.get("one").value_or("error"), "eins");
+    }
+
+    TEST(LRUCache, eraseFromNonEmptyCache) {
+        LRUCache<std::string, std::string> c(10);
+        c.upsert("one", "eins");
+        ASSERT_EQ(c.erase("one"), true);
+        ASSERT_EQ(c.size(), 0);
+        ASSERT_EQ(c.get("one").value_or("empty"), "empty");
+    }
+}

src/libutil/url.cc

@@ -4,6 +4,7 @@
 namespace nix {
 
 std::regex refRegex(refRegexS, std::regex::ECMAScript);
+std::regex badGitRefRegex(badGitRefRegexS, std::regex::ECMAScript);
 std::regex revRegex(revRegexS, std::regex::ECMAScript);
 std::regex flakeIdRegex(flakeIdRegexS, std::regex::ECMAScript);
 

src/libutil/url.hh

@@ -49,6 +49,12 @@ const static std::string pathRegex = "(?:" + segmentRegex + "(?:/" + segmentRege
 const static std::string refRegexS = "[a-zA-Z0-9][a-zA-Z0-9_.-]*"; // FIXME: check
 extern std::regex refRegex;
 
+// Instead of defining what a good Git Ref is, we define what a bad Git Ref is
+// This is because of the definition of a ref in refs.c in https://github.com/git/git
+// See tests/fetchGitRefs.sh for the full definition
+const static std::string badGitRefRegexS = "//|^[./]|/\\.|\\.\\.|[[:cntrl:][:space:]:?^~\[]|\\\\|\\*|\\.lock$|\\.lock/|@\\{|[/.]$|^@$|^$";
+extern std::regex badGitRefRegex;
+
 // A Git revision (a SHA-1 commit hash).
 const static std::string revRegexS = "[0-9a-fA-F]{40}";
 extern std::regex revRegex;

src/libutil/util.hh

@@ -459,8 +459,7 @@ string base64Encode(const string & s);
 string base64Decode(const string & s);
 
 
-/* Get a value for the specified key from an associate container, or a
+/* Get a value for the specified key from an associate container. */
-   default value if the key doesn't exist. */
 template <class T>
 std::optional<typename T::mapped_type> get(const T & map, const typename T::key_type & key)
 {

src/nix/add-to-store.cc

@@ -50,8 +50,10 @@ struct CmdAddToStore : MixDryRun, StoreCommand
         info.narSize = sink.s->size();
         info.ca = makeFixedOutputCA(FileIngestionMethod::Recursive, info.narHash);
 
-        if (!dryRun)
+        if (!dryRun) {
-            store->addToStore(info, sink.s);
+            auto source = StringSource { *sink.s };
+            store->addToStore(info, source);
+        }
 
         logger->stdout("%s", store->printStorePath(info.path));
     }

src/nix/develop.cc

@@ -110,7 +110,7 @@ StorePath getDerivationEnvironment(ref<Store> store, const StorePath & drvPath)
 
     auto builder = baseNameOf(drv.builder);
     if (builder != "bash")
-        throw Error("'nix dev-shell' only works on derivations that use 'bash' as their builder");
+        throw Error("'nix develop' only works on derivations that use 'bash' as their builder");
 
     auto getEnvShPath = store->addTextToStore("get-env.sh", getEnvSh, {});
 
@@ -233,11 +233,11 @@ struct Common : InstallableCommand, MixProfile
     }
 };
 
-struct CmdDevShell : Common, MixEnvironment
+struct CmdDevelop : Common, MixEnvironment
 {
     std::vector<std::string> command;
 
-    CmdDevShell()
+    CmdDevelop()
     {
         addFlag({
             .longName = "command",
@@ -261,15 +261,15 @@ struct CmdDevShell : Common, MixEnvironment
         return {
             Example{
                 "To get the build environment of GNU hello:",
-                "nix dev-shell nixpkgs.hello"
+                "nix develop nixpkgs.hello"
             },
             Example{
                 "To store the build environment in a profile:",
-                "nix dev-shell --profile /tmp/my-shell nixpkgs.hello"
+                "nix develop --profile /tmp/my-shell nixpkgs.hello"
             },
             Example{
                 "To use a build environment previously recorded in a profile:",
-                "nix dev-shell /tmp/my-shell"
+                "nix develop /tmp/my-shell"
             },
         };
     }
@@ -343,4 +343,4 @@ struct CmdPrintDevEnv : Common
 };
 
 static auto r1 = registerCommand<CmdPrintDevEnv>("print-dev-env");
-static auto r2 = registerCommand<CmdDevShell>("dev-shell");
+static auto r2 = registerCommand<CmdDevelop>("develop");

src/nix/local.mk

@@ -28,4 +28,4 @@ $(eval $(call install-symlink, $(bindir)/nix, $(libexecdir)/nix/build-remote))
 
 src/nix-env/user-env.cc: src/nix-env/buildenv.nix.gen.hh
 
-src/nix/dev-shell.cc: src/nix/get-env.sh.gen.hh
+src/nix/develop.cc: src/nix/get-env.sh.gen.hh

src/nix/main.cc

@@ -110,6 +110,8 @@ struct NixArgs : virtual MultiCommand, virtual MixCommonArgs
             .description = "consider all previously downloaded files out-of-date",
             .handler = {[&]() { refresh = true; }},
         });
+
+        deprecatedAliases.insert({"dev-shell", "develop"});
     }
 
     void printFlags(std::ostream & out) override

tests/fetchGitRefs.sh

@@ -0,0 +1,111 @@
+source common.sh
+
+if [[ -z $(type -p git) ]]; then
+    echo "Git not installed; skipping Git tests"
+    exit 99
+fi
+
+clearStore
+
+repo="$TEST_ROOT/git"
+
+rm -rf "$repo" "${repo}-tmp" "$TEST_HOME/.cache/nix"
+
+git init "$repo"
+git -C "$repo" config user.email "foobar@example.com"
+git -C "$repo" config user.name "Foobar"
+
+echo utrecht > "$repo"/hello
+git -C "$repo" add hello
+git -C "$repo" commit -m 'Bla1'
+
+path=$(nix eval --raw "(builtins.fetchGit { url = $repo; ref = \"master\"; }).outPath")
+
+# Test various combinations of ref names
+# (taken from the git project)
+
+# git help check-ref-format
+#       Git imposes the following rules on how references are named:
+#
+#        1. They can include slash / for hierarchical (directory) grouping, but no slash-separated component can begin with a dot .  or end with the sequence .lock.
+#        2. They must contain at least one /. This enforces the presence of a category like heads/, tags/ etc. but the actual names are not restricted. If the --allow-onelevel option is used, this rule is waived.
+#        3. They cannot have two consecutive dots ..  anywhere.
+#        4. They cannot have ASCII control characters (i.e. bytes whose values are lower than \040, or \177 DEL), space, tilde ~, caret ^, or colon : anywhere.
+#        5. They cannot have question-mark ?, asterisk *, or open bracket [ anywhere. See the --refspec-pattern option below for an exception to this rule.
+#        6. They cannot begin or end with a slash / or contain multiple consecutive slashes (see the --normalize option below for an exception to this rule)
+#        7. They cannot end with a dot ..
+#        8. They cannot contain a sequence @{.
+#        9. They cannot be the single character @.
+#       10. They cannot contain a \.
+
+valid_ref() {
+    { set +x; printf >&2 '\n>>>>>>>>>> valid_ref %s\b <<<<<<<<<<\n' $(printf %s "$1" | sed -n -e l); set -x; }
+    git check-ref-format --branch "$1" >/dev/null
+    git -C "$repo" branch "$1" master >/dev/null
+    path1=$(nix eval --raw "(builtins.fetchGit { url = $repo; ref = ''$1''; }).outPath")
+    [[ $path1 = $path ]]
+    git -C "$repo" branch -D "$1" >/dev/null
+}
+
+invalid_ref() {
+    { set +x; printf >&2 '\n>>>>>>>>>> invalid_ref %s\b <<<<<<<<<<\n' $(printf %s "$1" | sed -n -e l); set -x; }
+    # special case for a sole @:
+    # --branch @ will try to interpret @ as a branch reference and not fail. Thus we need --allow-onelevel
+    if [ "$1" = "@" ]; then
+        (! git check-ref-format --allow-onelevel "$1" >/dev/null 2>&1)
+    else
+        (! git check-ref-format --branch "$1" >/dev/null 2>&1)
+    fi
+    nix --debug eval --raw "(builtins.fetchGit { url = $repo; ref = ''$1''; }).outPath" 2>&1 | grep 'error: invalid Git branch/tag name' >/dev/null
+}
+
+
+valid_ref 'foox'
+valid_ref '1337'
+valid_ref 'foo.baz'
+valid_ref 'foo/bar/baz'
+valid_ref 'foo./bar'
+valid_ref 'heads/foo@bar'
+valid_ref "$(printf 'heads/fu\303\237')"
+valid_ref 'foo-bar-baz'
+valid_ref '$1'
+valid_ref 'foo.locke'
+
+invalid_ref 'refs///heads/foo'
+invalid_ref 'heads/foo/'
+invalid_ref '///heads/foo'
+invalid_ref '.foo'
+invalid_ref './foo'
+invalid_ref './foo/bar'
+invalid_ref 'foo/./bar'
+invalid_ref 'foo/bar/.'
+invalid_ref 'foo bar'
+invalid_ref 'foo?bar'
+invalid_ref 'foo^bar'
+invalid_ref 'foo~bar'
+invalid_ref 'foo:bar'
+invalid_ref 'foo[bar'
+invalid_ref 'foo/bar/.'
+invalid_ref '.refs/foo'
+invalid_ref 'refs/heads/foo.'
+invalid_ref 'heads/foo..bar'
+invalid_ref 'heads/foo?bar'
+invalid_ref 'heads/foo.lock'
+invalid_ref 'heads///foo.lock'
+invalid_ref 'foo.lock/bar'
+invalid_ref 'foo.lock///bar'
+invalid_ref 'heads/v@{ation'
+invalid_ref 'heads/foo\.ar' # should fail due to \
+invalid_ref 'heads/foo\bar' # should fail due to \
+invalid_ref "$(printf 'heads/foo\t')" # should fail because it has a TAB
+invalid_ref "$(printf 'heads/foo\177')"
+invalid_ref '@'
+
+invalid_ref 'foo/*'
+invalid_ref '*/foo'
+invalid_ref 'foo/*/bar'
+invalid_ref '*'
+invalid_ref 'foo/*/*'
+invalid_ref '*/foo/*'
+invalid_ref '/foo'
+invalid_ref ''

tests/local.mk

@@ -18,6 +18,7 @@ nix_tests = \
   nar-access.sh \
   structured-attrs.sh \
   fetchGit.sh \
+  fetchGitRefs.sh \
   fetchGitSubmodules.sh \
   fetchMercurial.sh \
   signing.sh \