Auto merge of #118412 - matthiaskrgr:rollup-ghzhti2, r=matthiaskrgr

Rollup of 6 pull requests

Successful merges:

 - #118193 (Add missing period in `std::process::Command` docs)
 - #118222 (unify read_to_end and io::copy impls for reading into a Vec)
 - #118323 (give dev-friendly error message for incorrect config profiles)
 - #118378 (Perform LTO optimisations with wasm-ld + -Clinker-plugin-lto)
 - #118399 (Clean dead codes in miri)
 - #118410 (update test for new LLVM 18 codegen)

r? `@ghost`
`@rustbot` modify labels: rollup

compiler/rustc_codegen_ssa/src/back/linker.rs

@@ -1308,6 +1308,8 @@ impl<'a> Linker for WasmLd<'a> {
     }
 
     fn optimize(&mut self) {
+        // The -O flag is, as of late 2023, only used for merging of strings and debuginfo, and
+        // only differentiates -O0 and -O1. It does not apply to LTO.
         self.cmd.arg(match self.sess.opts.optimize {
             OptLevel::No => "-O0",
             OptLevel::Less => "-O1",
@@ -1360,7 +1362,31 @@ impl<'a> Linker for WasmLd<'a> {
     fn subsystem(&mut self, _subsystem: &str) {}
 
     fn linker_plugin_lto(&mut self) {
-        // Do nothing for now
+        match self.sess.opts.cg.linker_plugin_lto {
+            LinkerPluginLto::Disabled => {
+                // Nothing to do
+            }
+            LinkerPluginLto::LinkerPluginAuto => {
+                self.push_linker_plugin_lto_args();
+            }
+            LinkerPluginLto::LinkerPlugin(_) => {
+                self.push_linker_plugin_lto_args();
+            }
+        }
+    }
+}
+
+impl<'a> WasmLd<'a> {
+    fn push_linker_plugin_lto_args(&mut self) {
+        let opt_level = match self.sess.opts.optimize {
+            config::OptLevel::No => "O0",
+            config::OptLevel::Less => "O1",
+            config::OptLevel::Default => "O2",
+            config::OptLevel::Aggressive => "O3",
+            // wasm-ld only handles integer LTO opt levels. Use O2
+            config::OptLevel::Size | config::OptLevel::SizeMin => "O2",
+        };
+        self.cmd.arg(&format!("--lto-{opt_level}"));
     }
 }
 

library/std/src/io/copy.rs

@@ -1,7 +1,6 @@
 use super::{BorrowedBuf, BufReader, BufWriter, Read, Result, Write, DEFAULT_BUF_SIZE};
 use crate::alloc::Allocator;
 use crate::cmp;
-use crate::cmp::min;
 use crate::collections::VecDeque;
 use crate::io::IoSlice;
 use crate::mem::MaybeUninit;
@@ -256,79 +255,17 @@ impl<I: Write + ?Sized> BufferedWriterSpec for BufWriter<I> {
     }
 }
 
-impl<A: Allocator> BufferedWriterSpec for Vec<u8, A> {
+impl BufferedWriterSpec for Vec<u8> {
     fn buffer_size(&self) -> usize {
         cmp::max(DEFAULT_BUF_SIZE, self.capacity() - self.len())
     }
 
     fn copy_from<R: Read + ?Sized>(&mut self, reader: &mut R) -> Result<u64> {
-        let mut bytes = 0;
+        reader.read_to_end(self).map(|bytes| u64::try_from(bytes).expect("usize overflowed u64"))
-
-        // avoid inflating empty/small vecs before we have determined that there's anything to read
-        if self.capacity() < DEFAULT_BUF_SIZE {
-            let stack_read_limit = DEFAULT_BUF_SIZE as u64;
-            bytes = stack_buffer_copy(&mut reader.take(stack_read_limit), self)?;
-            // fewer bytes than requested -> EOF reached
-            if bytes < stack_read_limit {
-                return Ok(bytes);
-            }
-        }
-
-        // don't immediately offer the vec's whole spare capacity, otherwise
-        // we might have to fully initialize it if the reader doesn't have a custom read_buf() impl
-        let mut max_read_size = DEFAULT_BUF_SIZE;
-
-        loop {
-            self.reserve(DEFAULT_BUF_SIZE);
-            let mut initialized_spare_capacity = 0;
-
-            loop {
-                let buf = self.spare_capacity_mut();
-                let read_size = min(max_read_size, buf.len());
-                let mut buf = BorrowedBuf::from(&mut buf[..read_size]);
-                // SAFETY: init is either 0 or the init_len from the previous iteration.
-                unsafe {
-                    buf.set_init(initialized_spare_capacity);
-                }
-                match reader.read_buf(buf.unfilled()) {
-                    Ok(()) => {
-                        let bytes_read = buf.len();
-
-                        // EOF
-                        if bytes_read == 0 {
-                            return Ok(bytes);
-                        }
-
-                        // the reader is returning short reads but it doesn't call ensure_init()
-                        if buf.init_len() < buf.capacity() {
-                            max_read_size = usize::MAX;
-                        }
-                        // the reader hasn't returned short reads so far
-                        if bytes_read == buf.capacity() {
-                            max_read_size *= 2;
-                        }
-
-                        initialized_spare_capacity = buf.init_len() - bytes_read;
-                        bytes += bytes_read as u64;
-                        // SAFETY: BorrowedBuf guarantees all of its filled bytes are init
-                        // and the number of read bytes can't exceed the spare capacity since
-                        // that's what the buffer is borrowing from.
-                        unsafe { self.set_len(self.len() + bytes_read) };
-
-                        // spare capacity full, reserve more
-                        if self.len() == self.capacity() {
-                            break;
-                        }
-                    }
-                    Err(e) if e.is_interrupted() => continue,
-                    Err(e) => return Err(e),
-                }
-            }
-        }
     }
 }
 
-fn stack_buffer_copy<R: Read + ?Sized, W: Write + ?Sized>(
+pub fn stack_buffer_copy<R: Read + ?Sized, W: Write + ?Sized>(
     reader: &mut R,
     writer: &mut W,
 ) -> Result<u64> {

library/std/src/io/copy/tests.rs

@@ -82,13 +82,16 @@ fn copy_specializes_bufreader() {
 
 #[test]
 fn copy_specializes_to_vec() {
-    let cap = 123456;
+    let cap = DEFAULT_BUF_SIZE * 10;
-    let mut source = ShortReader { cap, observed_buffer: 0, read_size: 1337 };
+    let mut source = ShortReader { cap, observed_buffer: 0, read_size: DEFAULT_BUF_SIZE };
     let mut sink = Vec::new();
-    assert_eq!(cap as u64, io::copy(&mut source, &mut sink).unwrap());
+    let copied = io::copy(&mut source, &mut sink).unwrap();
+    assert_eq!(cap as u64, copied);
+    assert_eq!(sink.len() as u64, copied);
     assert!(
         source.observed_buffer > DEFAULT_BUF_SIZE,
-        "expected a large buffer to be provided to the reader"
+        "expected a large buffer to be provided to the reader, got {}",
+        source.observed_buffer
     );
 }
 

library/std/src/io/mod.rs

@@ -397,12 +397,16 @@ where
     }
 }
 
-// This uses an adaptive system to extend the vector when it fills. We want to
+// Here we must serve many masters with conflicting goals:
-// avoid paying to allocate and zero a huge chunk of memory if the reader only
+//
-// has 4 bytes while still making large reads if the reader does have a ton
+// - avoid allocating unless necessary
-// of data to return. Simply tacking on an extra DEFAULT_BUF_SIZE space every
+// - avoid overallocating if we know the exact size (#89165)
-// time is 4,500 times (!) slower than a default reservation size of 32 if the
+// - avoid passing large buffers to readers that always initialize the free capacity if they perform short reads (#23815, #23820)
-// reader has a very small amount of data to return.
+// - pass large buffers to readers that do not initialize the spare capacity. this can amortize per-call overheads
+// - and finally pass not-too-small and not-too-large buffers to Windows read APIs because they manage to suffer from both problems
+//   at the same time, i.e. small reads suffer from syscall overhead, all reads incur initialization cost
+//   proportional to buffer size (#110650)
+//
 pub(crate) fn default_read_to_end<R: Read + ?Sized>(
     r: &mut R,
     buf: &mut Vec<u8>,
@@ -412,20 +416,58 @@ pub(crate) fn default_read_to_end<R: Read + ?Sized>(
     let start_cap = buf.capacity();
     // Optionally limit the maximum bytes read on each iteration.
     // This adds an arbitrary fiddle factor to allow for more data than we expect.
-    let max_read_size =
+    let mut max_read_size = size_hint
-        size_hint.and_then(|s| s.checked_add(1024)?.checked_next_multiple_of(DEFAULT_BUF_SIZE));
+        .and_then(|s| s.checked_add(1024)?.checked_next_multiple_of(DEFAULT_BUF_SIZE))
+        .unwrap_or(DEFAULT_BUF_SIZE);
 
     let mut initialized = 0; // Extra initialized bytes from previous loop iteration
+
+    const PROBE_SIZE: usize = 32;
+
+    fn small_probe_read<R: Read + ?Sized>(r: &mut R, buf: &mut Vec<u8>) -> Result<usize> {
+        let mut probe = [0u8; PROBE_SIZE];
+
+        loop {
+            match r.read(&mut probe) {
+                Ok(n) => {
+                    buf.extend_from_slice(&probe[..n]);
+                    return Ok(n);
+                }
+                Err(ref e) if e.is_interrupted() => continue,
+                Err(e) => return Err(e),
+            }
+        }
+    }
+
+    // avoid inflating empty/small vecs before we have determined that there's anything to read
+    if (size_hint.is_none() || size_hint == Some(0)) && buf.capacity() - buf.len() < PROBE_SIZE {
+        let read = small_probe_read(r, buf)?;
+
+        if read == 0 {
+            return Ok(0);
+        }
+    }
+
     loop {
+        if buf.len() == buf.capacity() && buf.capacity() == start_cap {
+            // The buffer might be an exact fit. Let's read into a probe buffer
+            // and see if it returns `Ok(0)`. If so, we've avoided an
+            // unnecessary doubling of the capacity. But if not, append the
+            // probe buffer to the primary buffer and let its capacity grow.
+            let read = small_probe_read(r, buf)?;
+
+            if read == 0 {
+                return Ok(buf.len() - start_len);
+            }
+        }
+
         if buf.len() == buf.capacity() {
-            buf.reserve(32); // buf is full, need more space
+            buf.reserve(PROBE_SIZE); // buf is full, need more space
         }
 
         let mut spare = buf.spare_capacity_mut();
-        if let Some(size) = max_read_size {
+        let buf_len = cmp::min(spare.len(), max_read_size);
-            let len = cmp::min(spare.len(), size);
+        spare = &mut spare[..buf_len];
-            spare = &mut spare[..len]
-        }
         let mut read_buf: BorrowedBuf<'_> = spare.into();
 
         // SAFETY: These bytes were initialized but not filled in the previous loop
@@ -434,42 +476,44 @@ pub(crate) fn default_read_to_end<R: Read + ?Sized>(
         }
 
         let mut cursor = read_buf.unfilled();
-        match r.read_buf(cursor.reborrow()) {
+        loop {
-            Ok(()) => {}
+            match r.read_buf(cursor.reborrow()) {
-            Err(e) if e.is_interrupted() => continue,
+                Ok(()) => break,
-            Err(e) => return Err(e),
+                Err(e) if e.is_interrupted() => continue,
+                Err(e) => return Err(e),
+            }
         }
 
-        if cursor.written() == 0 {
+        let unfilled_but_initialized = cursor.init_ref().len();
+        let bytes_read = cursor.written();
+        let was_fully_initialized = read_buf.init_len() == buf_len;
+
+        if bytes_read == 0 {
             return Ok(buf.len() - start_len);
         }
 
         // store how much was initialized but not filled
-        initialized = cursor.init_ref().len();
+        initialized = unfilled_but_initialized;
 
         // SAFETY: BorrowedBuf's invariants mean this much memory is initialized.
         unsafe {
-            let new_len = read_buf.filled().len() + buf.len();
+            let new_len = bytes_read + buf.len();
             buf.set_len(new_len);
         }
 
-        if buf.len() == buf.capacity() && buf.capacity() == start_cap {
+        // Use heuristics to determine the max read size if no initial size hint was provided
-            // The buffer might be an exact fit. Let's read into a probe buffer
+        if size_hint.is_none() {
-            // and see if it returns `Ok(0)`. If so, we've avoided an
+            // The reader is returning short reads but it doesn't call ensure_init().
-            // unnecessary doubling of the capacity. But if not, append the
+            // In that case we no longer need to restrict read sizes to avoid
-            // probe buffer to the primary buffer and let its capacity grow.
+            // initialization costs.
-            let mut probe = [0u8; 32];
+            if !was_fully_initialized {
+                max_read_size = usize::MAX;
+            }
 
-            loop {
+            // we have passed a larger buffer than previously and the
-                match r.read(&mut probe) {
+            // reader still hasn't returned a short read
-                    Ok(0) => return Ok(buf.len() - start_len),
+            if buf_len >= max_read_size && bytes_read == buf_len {
-                    Ok(n) => {
+                max_read_size = max_read_size.saturating_mul(2);
-                        buf.extend_from_slice(&probe[..n]);
-                        break;
-                    }
-                    Err(ref e) if e.is_interrupted() => continue,
-                    Err(e) => return Err(e),
-                }
             }
         }
     }

library/std/src/process.rs

@@ -1108,7 +1108,7 @@ impl fmt::Debug for Command {
     ///
     /// The default format approximates a shell invocation of the program along with its
     /// arguments. It does not include most of the other command properties. The output is not guaranteed to work
-    /// (e.g. due to lack of shell-escaping or differences in path resolution)
+    /// (e.g. due to lack of shell-escaping or differences in path resolution).
     /// On some platforms you can use [the alternate syntax] to show more fields.
     ///
     /// Note that the debug implementation is platform-specific.

src/bootstrap/bootstrap.py

@@ -1083,6 +1083,11 @@ def bootstrap(args):
         include_file = 'config.{}.toml'.format(profile_aliases.get(profile) or profile)
         include_dir = os.path.join(rust_root, 'src', 'bootstrap', 'defaults')
         include_path = os.path.join(include_dir, include_file)
+
+        if not os.path.exists(include_path):
+            raise Exception("Unrecognized config profile '{}'. Check src/bootstrap/defaults"
+            " for available options.".format(profile))
+
         # HACK: This works because `self.get_toml()` returns the first match it finds for a
         # specific key, so appending our defaults at the end allows the user to override them
         with open(include_path) as included_toml:

src/tools/miri/src/shims/unix/fs.rs

@@ -66,11 +66,6 @@ pub trait FileDescriptor: std::fmt::Debug + Any {
     fn is_tty(&self, _communicate_allowed: bool) -> bool {
         false
     }
-
-    #[cfg(unix)]
-    fn as_unix_host_fd(&self) -> Option<i32> {
-        None
-    }
 }
 
 impl dyn FileDescriptor {
@@ -150,12 +145,6 @@ impl FileDescriptor for FileHandle {
         Ok(Box::new(FileHandle { file: duplicated, writable: self.writable }))
     }
 
-    #[cfg(unix)]
-    fn as_unix_host_fd(&self) -> Option<i32> {
-        use std::os::unix::io::AsRawFd;
-        Some(self.file.as_raw_fd())
-    }
-
     fn is_tty(&self, communicate_allowed: bool) -> bool {
         communicate_allowed && self.file.is_terminal()
     }
@@ -183,11 +172,6 @@ impl FileDescriptor for io::Stdin {
         Ok(Box::new(io::stdin()))
     }
 
-    #[cfg(unix)]
-    fn as_unix_host_fd(&self) -> Option<i32> {
-        Some(libc::STDIN_FILENO)
-    }
-
     fn is_tty(&self, communicate_allowed: bool) -> bool {
         communicate_allowed && self.is_terminal()
     }
@@ -220,11 +204,6 @@ impl FileDescriptor for io::Stdout {
         Ok(Box::new(io::stdout()))
     }
 
-    #[cfg(unix)]
-    fn as_unix_host_fd(&self) -> Option<i32> {
-        Some(libc::STDOUT_FILENO)
-    }
-
     fn is_tty(&self, communicate_allowed: bool) -> bool {
         communicate_allowed && self.is_terminal()
     }
@@ -250,11 +229,6 @@ impl FileDescriptor for io::Stderr {
         Ok(Box::new(io::stderr()))
     }
 
-    #[cfg(unix)]
-    fn as_unix_host_fd(&self) -> Option<i32> {
-        Some(libc::STDERR_FILENO)
-    }
-
     fn is_tty(&self, communicate_allowed: bool) -> bool {
         communicate_allowed && self.is_terminal()
     }

tests/codegen/ascii-char.rs

@@ -14,7 +14,7 @@ pub fn unwrap_digit_from_remainder(v: u32) -> AsciiChar {
 
     // CHECK: %[[R:.+]] = urem i32 %v, 10
     // CHECK-NEXT: %[[T:.+]] = trunc i32 %[[R]] to i8
-    // CHECK-NEXT: %[[D:.+]] = or i8 %[[T]], 48
+    // CHECK-NEXT: %[[D:.+]] = or{{( disjoint)?}} i8 %[[T]], 48
     // CHECK-NEXT: ret i8 %[[D]]
 
     // CHECK-NOT: icmp