tree -> index diff for status

.github/workflows/msrv.yml

@@ -25,6 +25,6 @@ jobs:
       - uses: actions/checkout@v4
       - uses: extractions/setup-just@v2
       - run: |
-          rustup toolchain install $rust_version --profile minimal --no-self-update
-          rustup default $rust_version
+          rustup toolchain install ${{ env.rust_version }} --profile minimal --no-self-update
+          rustup default ${{ env.rust_version }}
       - run: just ci-check-msrv

gitoxide-core/src/repository/odb.rs

@@ -1,4 +1,5 @@
 use std::io;
+use std::sync::atomic::Ordering;
 
 use anyhow::bail;
 
@@ -50,6 +51,8 @@ pub mod statistics {
     pub struct Options {
         pub format: OutputFormat,
         pub thread_limit: Option<usize>,
+        /// A debug-flag that triggers looking up the headers of all objects again, but without indices preloaded
+        pub extra_header_lookup: bool,
     }
 }
 
@@ -59,7 +62,11 @@ pub fn statistics(
     mut progress: impl gix::Progress,
     out: impl io::Write,
     mut err: impl io::Write,
-    statistics::Options { format, thread_limit }: statistics::Options,
+    statistics::Options {
+        format,
+        thread_limit,
+        extra_header_lookup,
+    }: statistics::Options,
 ) -> anyhow::Result<()> {
     use bytesize::ByteSize;
     use gix::odb::{find, HeaderExt};
@@ -76,6 +83,10 @@ pub fn statistics(
     #[cfg_attr(feature = "serde", derive(serde::Serialize))]
     #[derive(Default)]
     struct Statistics {
+        /// All objects that were used to produce these statistics.
+        /// Only `Some` if we are doing an extra round of header queries on a repository without loaded indices.
+        #[cfg_attr(feature = "serde", serde(skip_serializing))]
+        ids: Option<Vec<gix::ObjectId>>,
         total_objects: usize,
         loose_objects: usize,
         packed_objects: usize,
@@ -135,14 +146,17 @@ pub fn statistics(
     }
 
     impl gix::parallel::Reduce for Reduce {
-        type Input = Result<Vec<gix::odb::find::Header>, anyhow::Error>;
+        type Input = Result<Vec<(gix::ObjectId, gix::odb::find::Header)>, anyhow::Error>;
         type FeedProduce = ();
         type Output = Statistics;
         type Error = anyhow::Error;
 
         fn feed(&mut self, items: Self::Input) -> Result<Self::FeedProduce, Self::Error> {
-            for item in items? {
+            for (id, item) in items? {
                 self.stats.consume(item);
+                if let Some(ids) = self.stats.ids.as_mut() {
+                    ids.push(id);
+                }
             }
             Ok(())
         }
@@ -154,9 +168,9 @@ pub fn statistics(
     }
 
     let cancelled = || anyhow::anyhow!("Cancelled by user");
-    let object_ids = repo.objects.store_ref().iter()?.filter_map(Result::ok);
+    let object_ids = repo.objects.iter()?.filter_map(Result::ok);
     let chunk_size = 1_000;
-    let stats = if gix::parallel::num_threads(thread_limit) > 1 {
+    let mut stats = if gix::parallel::num_threads(thread_limit) > 1 {
         gix::parallel::in_parallel(
             gix::interrupt::Iter::new(
                 gix::features::iter::Chunks {
@@ -166,19 +180,30 @@ pub fn statistics(
                 cancelled,
             ),
             thread_limit,
-            move |_| (repo.objects.clone().into_inner(), counter),
+            {
+                let objects = repo.objects.clone();
+                move |_| (objects.clone().into_inner(), counter)
+            },
             |ids, (handle, counter)| {
                 let ids = ids?;
-                counter.fetch_add(ids.len(), std::sync::atomic::Ordering::Relaxed);
+                counter.fetch_add(ids.len(), Ordering::Relaxed);
                 let out = ids
                     .into_iter()
-                    .map(|id| handle.header(id))
+                    .map(|id| handle.header(id).map(|hdr| (id, hdr)))
                     .collect::<Result<Vec<_>, _>>()?;
                 Ok(out)
             },
-            Reduce::default(),
+            Reduce {
+                stats: Statistics {
+                    ids: extra_header_lookup.then(Vec::new),
+                    ..Default::default()
+                },
+            },
         )?
     } else {
+        if extra_header_lookup {
+            bail!("extra-header-lookup is only meaningful in threaded mode");
+        }
         let mut stats = Statistics::default();
 
         for (count, id) in object_ids.enumerate() {
@@ -193,6 +218,39 @@ pub fn statistics(
 
     progress.show_throughput(start);
 
+    if let Some(mut ids) = stats.ids.take() {
+        // Critical to re-open the repo to assure we don't have any ODB state and start fresh.
+        let start = std::time::Instant::now();
+        let repo = gix::open_opts(repo.git_dir(), repo.open_options().to_owned())?;
+        progress.set_name("re-counting".into());
+        progress.init(Some(ids.len()), gix::progress::count("objects"));
+        let counter = progress.counter();
+        counter.store(0, Ordering::Relaxed);
+        let errors = gix::parallel::in_parallel_with_slice(
+            &mut ids,
+            thread_limit,
+            {
+                let objects = repo.objects.clone();
+                move |_| (objects.clone().into_inner(), counter, false)
+            },
+            |id, (odb, counter, has_error), _threads_left, _stop_everything| -> anyhow::Result<()> {
+                counter.fetch_add(1, Ordering::Relaxed);
+                if let Err(_err) = odb.header(id) {
+                    *has_error = true;
+                    gix::trace::error!(err = ?_err, "Object that is known to be present wasn't found");
+                }
+                Ok(())
+            },
+            || Some(std::time::Duration::from_millis(100)),
+            |(_, _, has_error)| has_error,
+        )?;
+
+        progress.show_throughput(start);
+        if errors.contains(&true) {
+            bail!("At least one object couldn't be looked up even though it must exist");
+        }
+    }
+
     #[cfg(feature = "serde")]
     {
         serde_json::to_writer_pretty(out, &stats)?;

gix-config/tests/mem.rs

@@ -4,7 +4,7 @@ use std::alloc;
 use std::time::Instant;
 
 #[global_allocator]
-static ALLOCATOR: Cap<alloc::System> = Cap::new(alloc::System, usize::max_value());
+static ALLOCATOR: Cap<alloc::System> = Cap::new(alloc::System, usize::MAX);
 
 #[test]
 fn usage() {

gix-date/src/time/write.rs

@@ -5,6 +5,10 @@ use crate::{time::Sign, Time};
 /// Serialization with standard `git` format
 impl Time {
     /// Serialize this instance into memory, similar to what [`write_to()`][Self::write_to()] would do with arbitrary `Write` implementations.
+    ///
+    /// # Panics
+    ///
+    /// If the underlying call fails as this instance can't be represented, typically due to an invalid offset.
     pub fn to_bstring(&self) -> BString {
         let mut buf = Vec::with_capacity(64);
         self.write_to(&mut buf).expect("write to memory cannot fail");
@@ -13,6 +17,18 @@ impl Time {
 
     /// Serialize this instance to `out` in a format suitable for use in header fields of serialized git commits or tags.
     pub fn write_to(&self, out: &mut dyn std::io::Write) -> std::io::Result<()> {
+        const SECONDS_PER_HOUR: u32 = 60 * 60;
+        let offset = self.offset.unsigned_abs();
+        let hours = offset / SECONDS_PER_HOUR;
+        let minutes = (offset - (hours * SECONDS_PER_HOUR)) / 60;
+
+        if hours > 99 {
+            return Err(std::io::Error::new(
+                std::io::ErrorKind::Other,
+                "Cannot represent offsets larger than +-9900",
+            ));
+        }
+
         let mut itoa = itoa::Buffer::new();
         out.write_all(itoa.format(self.seconds).as_bytes())?;
         out.write_all(b" ")?;
@@ -23,12 +39,6 @@ impl Time {
 
         const ZERO: &[u8; 1] = b"0";
 
-        const SECONDS_PER_HOUR: u32 = 60 * 60;
-        let offset = self.offset.unsigned_abs();
-        let hours = offset / SECONDS_PER_HOUR;
-        assert!(hours < 25, "offset is more than a day: {hours}");
-        let minutes = (offset - (hours * SECONDS_PER_HOUR)) / 60;
-
         if hours < 10 {
             out.write_all(ZERO)?;
         }

gix-date/tests/time/mod.rs

@@ -1,5 +1,4 @@
-use bstr::ByteSlice;
-use gix_date::{time::Sign, SecondsSinceUnixEpoch, Time};
+use gix_date::Time;
 
 mod baseline;
 mod format;
@@ -32,57 +31,114 @@ fn is_set() {
     .is_set());
 }
 
-#[test]
-fn write_to() -> Result<(), Box<dyn std::error::Error>> {
-    for (time, expected) in [
-        (
-            Time {
-                seconds: SecondsSinceUnixEpoch::MAX,
-                offset: 0,
-                sign: Sign::Minus,
-            },
-            "9223372036854775807 -0000",
-        ),
-        (
-            Time {
-                seconds: SecondsSinceUnixEpoch::MIN,
-                offset: 0,
-                sign: Sign::Minus,
-            },
-            "-9223372036854775808 -0000",
-        ),
-        (
-            Time {
-                seconds: 500,
-                offset: 9000,
-                sign: Sign::Plus,
-            },
-            "500 +0230",
-        ),
-        (
-            Time {
-                seconds: 189009009,
-                offset: -36000,
-                sign: Sign::Minus,
-            },
-            "189009009 -1000",
-        ),
-        (
-            Time {
-                seconds: 0,
-                offset: 0,
-                sign: Sign::Minus,
-            },
-            "0 -0000",
-        ),
-    ] {
-        let mut output = Vec::new();
-        time.write_to(&mut output)?;
-        assert_eq!(output.as_bstr(), expected);
-        assert_eq!(time.size(), output.len());
+mod write_to {
+    use bstr::ByteSlice;
+    use gix_date::time::Sign;
+    use gix_date::{SecondsSinceUnixEpoch, Time};
+
+    #[test]
+    fn invalid() {
+        let time = Time {
+            seconds: 0,
+            offset: (100 * 60 * 60) + 30 * 60,
+            sign: Sign::Plus,
+        };
+        let err = time.write_to(&mut Vec::new()).unwrap_err();
+        assert_eq!(err.to_string(), "Cannot represent offsets larger than +-9900");
+    }
 
-        let actual = gix_date::parse(&output.as_bstr().to_string(), None).expect("round-trippable");
-        assert_eq!(time, actual);
+    #[test]
+    fn valid_roundtrips() -> Result<(), Box<dyn std::error::Error>> {
+        for (time, expected) in [
+            (
+                Time {
+                    seconds: SecondsSinceUnixEpoch::MAX,
+                    offset: 0,
+                    sign: Sign::Minus,
+                },
+                "9223372036854775807 -0000",
+            ),
+            (
+                Time {
+                    seconds: SecondsSinceUnixEpoch::MIN,
+                    offset: 0,
+                    sign: Sign::Minus,
+                },
+                "-9223372036854775808 -0000",
+            ),
+            (
+                Time {
+                    seconds: 500,
+                    offset: 9000,
+                    sign: Sign::Plus,
+                },
+                "500 +0230",
+            ),
+            (
+                Time {
+                    seconds: 189009009,
+                    offset: -36000,
+                    sign: Sign::Minus,
+                },
+                "189009009 -1000",
+            ),
+            (
+                Time {
+                    seconds: 0,
+                    offset: 0,
+                    sign: Sign::Minus,
+                },
+                "0 -0000",
+            ),
+            (
+                Time {
+                    seconds: 0,
+                    offset: -24 * 60 * 60,
+                    sign: Sign::Minus,
+                },
+                "0 -2400",
+            ),
+            (
+                Time {
+                    seconds: 0,
+                    offset: 24 * 60 * 60,
+                    sign: Sign::Plus,
+                },
+                "0 +2400",
+            ),
+            (
+                Time {
+                    seconds: 0,
+                    offset: (25 * 60 * 60) + 30 * 60,
+                    sign: Sign::Plus,
+                },
+                "0 +2530",
+            ),
+            (
+                Time {
+                    seconds: 0,
+                    offset: (-25 * 60 * 60) - 30 * 60,
+                    sign: Sign::Minus,
+                },
+                "0 -2530",
+            ),
+            (
+                Time {
+                    seconds: 0,
+                    offset: (99 * 60 * 60) + 59 * 60,
+                    sign: Sign::Plus,
+                },
+                "0 +9959",
+            ),
+        ] {
+            let mut output = Vec::new();
+            time.write_to(&mut output)?;
+            assert_eq!(output.as_bstr(), expected);
+            assert_eq!(time.size(), output.len());
+
+            let actual = gix_date::parse(&output.as_bstr().to_string(), None).expect("round-trippable");
+            assert_eq!(time, actual);
+        }
+        Ok(())
     }
-    Ok(())
 }