od: make GNU test od.pl pass (#9334)

* feat: Add support for long double floating-point numbers and refine general float formatting.

* feat: Enhance `od` error reporting for file I/O, width, and offset parsing, including overflow detection and input validation.

* feat: Improve long double parsing by converting f128 to f64, enhance overflow error reporting with `libc::ERANGE`, and prevent final offset printing on input errors.

* style: Apply minor formatting adjustments across the `od` module.

* refactor: simplify float formatting logic and update string handling syntax

* fix: Correct float formatting logic to use decimal for numbers within range and exponential otherwise.

* refactor(test): use helper function in test_calculate_alignment

Replace repetitive assert_eq! calls with a new assert_alignment helper
to improve test readability and reduce code duplication. The helper
encapsulates alignment checks for OutputInfo::calculate_alignment,
making tests clearer and easier to maintain.

* feat(cspell): add ERANGE to jargon wordlist

Added "ERANGE" to the dictionary to prevent spell checker flagging it as a misspelling, as it's a valid errno constant from C libraries.

* feat(od): improve width error handling and subnormal float output

Refactor width option parsing in OdOptions to use i18n-compatible error messages via translate! macro, consolidating redundant error branches for better maintainability. Enhance float formatting for f16 and bf16 by introducing format_binary16_like helper to properly display subnormal values with exponential notation, removing the obsolete format_float_simple function and adding subnormal detection functions for accurate representation in od's output.

* refactor(od): simplify format_item_bf16 by removing redundant variable

Remove unnecessary `value` variable in `format_item_bf16` function, eliminating
a redundant cast and inline `f` directly for clarity and minor efficiency gain.

* fix(od): standardize option names in error messages

Remove hardcoded "--" prefixes from localization strings in en-US.ftl and fr-FR.ftl, replacing with a computed display name that includes "--" and optionally the short form (e.g., "--option" or "--option, -s"). Update parse_bytes_option and read_bytes functions to pass an option_display_name, enabling consistent error message formatting across localizations. Add validation to reject zero width values as invalid arguments. Improves user experience by providing clearer, more consistent option references in error outputs.

* refactor: condense format! macro in format_item_bf16 for readability

Removed unnecessary line breaks in the format! expression, keeping the code more concise while maintaining functionality. This improves code style in the float printing module.

* fix(od): add external quoting for filenames in error messages

The MultifileReader now uses `fname.maybe_quote().external(true)` when displaying permission and I/O errors, ensuring filenames are properly quoted for user-facing output (e.g., handling special characters that might confuse shells). This prevents potential issues with filename display in error logs.

* refactor(od): Rename f128_to_f64 to u128_to_f64 for clarity

Renamed the function in input_decoder.rs from f128_to_f64 to u128_to_f64
to accurately reflect its purpose of converting u128 integer bits to f64,
improving code readability and reducing potential confusion over float types.

* refactor(od): simplify error handling in OdOptions using combinators

Use map_err and the try operator to replace a verbose match statement,
making the code more concise and idiomatic Rust. This improves readability
without altering functionality.

* Update src/uu/od/src/od.rs

Co-authored-by: Daniel Hofstetter <[email protected]>

* Update src/uu/od/src/od.rs

Co-authored-by: Daniel Hofstetter <[email protected]>

* Update src/uu/od/src/parse_inputs.rs

Co-authored-by: Daniel Hofstetter <[email protected]>

* Update src/uu/od/src/od.rs

Co-authored-by: Daniel Hofstetter <[email protected]>

* Update src/uu/od/src/parse_inputs.rs

Co-authored-by: Daniel Hofstetter <[email protected]>

* Update src/uu/od/src/parse_inputs.rs

Co-authored-by: Daniel Hofstetter <[email protected]>

* refactor(od): remove leaking from translated error messages in parse_offset_operand

Eliminated use of `.leak()` and unnecessary `.to_string()` calls on translated error strings in the `parse_offset_operand` function. This simplifies error handling, improves memory safety by avoiding intentional leaks, and makes the code cleaner without functional changes.

---------

Co-authored-by: Daniel Hofstetter <[email protected]>

Author: mattsu2020

.vscode/cspell.dictionaries/jargon.wordlist.txt

@@ -41,6 +41,7 @@ duplicative
 dsync
 endianness
 enqueue
+ERANGE
 errored
 executable
 executables

Cargo.lock

@@ -23,6 +23,7 @@ clap = { workspace = true }
 half = { workspace = true }
 uucore = { workspace = true, features = ["parser"] }
 fluent = { workspace = true }
+libc.workspace = true
 
 [[bin]]
 name = "od"

src/uu/od/Cargo.toml

@@ -55,9 +55,10 @@ od-error-invalid-offset = invalid offset: {$offset}
 od-error-invalid-label = invalid label: {$label}
 od-error-too-many-inputs = too many inputs after --traditional: {$input}
 od-error-parse-failed = parse failed
-od-error-invalid-suffix = invalid suffix in --{$option} argument {$value}
-od-error-invalid-argument = invalid --{$option} argument {$value}
-od-error-argument-too-large = --{$option} argument {$value} too large
+od-error-overflow = Numerical result out of range
+od-error-invalid-suffix = invalid suffix in {$option} argument {$value}
+od-error-invalid-argument = invalid {$option} argument {$value}
+od-error-argument-too-large = {$option} argument {$value} too large
 od-error-skip-past-end = tried to skip past end of input
 
 # Help messages

src/uu/od/locales/en-US.ftl

@@ -56,9 +56,9 @@ od-error-invalid-offset = décalage invalide : {$offset}
 od-error-invalid-label = étiquette invalide : {$label}
 od-error-too-many-inputs = trop d'entrées après --traditional : {$input}
 od-error-parse-failed = échec de l'analyse
-od-error-invalid-suffix = suffixe invalide dans l'argument --{$option} {$value}
-od-error-invalid-argument = argument --{$option} invalide {$value}
-od-error-argument-too-large = argument --{$option} {$value} trop grand
+od-error-invalid-suffix = suffixe invalide dans l'argument {$option} {$value}
+od-error-invalid-argument = argument {$option} invalide {$value}
+od-error-argument-too-large = argument {$option} {$value} trop grand
 od-error-skip-past-end = tentative d'ignorer au-delà de la fin de l'entrée
 
 # Messages d'aide

src/uu/od/locales/fr-FR.ftl

@@ -52,5 +52,6 @@ gen_byte_order_ops! {
     read_i32, write_i32 -> i32,
     read_i64, write_i64 -> i64,
     read_f32, write_f32 -> f32,
-    read_f64, write_f64 -> f64
+    read_f64, write_f64 -> f64,
+    read_u128, write_u128 -> u128
 }

src/uu/od/src/byteorder_io.rs

@@ -12,6 +12,7 @@ use std::fmt;
 pub enum FormatWriter {
     IntWriter(fn(u64) -> String),
     FloatWriter(fn(f64) -> String),
+    LongDoubleWriter(fn(f64) -> String), // On most platforms, long double is f64 or emulated
     BFloatWriter(fn(f64) -> String),
     MultibyteWriter(fn(&[u8]) -> String),
 }
@@ -27,6 +28,10 @@ impl fmt::Debug for FormatWriter {
                 f.write_str("FloatWriter:")?;
                 fmt::Pointer::fmt(p, f)
             }
+            Self::LongDoubleWriter(ref p) => {
+                f.write_str("LongDoubleWriter:")?;
+                fmt::Pointer::fmt(p, f)
+            }
             Self::BFloatWriter(ref p) => {
                 f.write_str("BFloatWriter:")?;
                 fmt::Pointer::fmt(p, f)

src/uu/od/src/formatter_item_info.rs

@@ -3,7 +3,7 @@
 // For the full copyright and license information, please view the LICENSE
 // file that was distributed with this source code.
 
-// spell-checker:ignore bfloat multifile
+// spell-checker:ignore bfloat multifile mant
 
 use half::{bf16, f16};
 use std::io;
@@ -165,6 +165,61 @@ impl MemoryDecoder<'_> {
         let val = f32::from(bf16::from_bits(bits));
         f64::from(val)
     }
+
+    /// Returns a long double from the internal buffer at position `start`.
+    /// We read 16 bytes as u128 (respecting endianness) and convert to f64.
+    /// This ensures that endianness swapping works correctly even if we lose precision.
+    pub fn read_long_double(&self, start: usize) -> f64 {
+        let bits = self.byte_order.read_u128(&self.data[start..start + 16]);
+        u128_to_f64(bits)
+    }
+}
+
+fn u128_to_f64(u: u128) -> f64 {
+    let sign = (u >> 127) as u64;
+    let exp = ((u >> 112) & 0x7FFF) as u64;
+    let mant = u & ((1 << 112) - 1);
+
+    if exp == 0x7FFF {
+        // Infinity or NaN
+        if mant == 0 {
+            if sign == 0 {
+                f64::INFINITY
+            } else {
+                f64::NEG_INFINITY
+            }
+        } else {
+            f64::NAN
+        }
+    } else if exp == 0 {
+        // Subnormal or zero
+        if mant == 0 {
+            if sign == 0 { 0.0 } else { -0.0 }
+        } else {
+            // Subnormal f128 is too small for f64, flush to zero
+            if sign == 0 { 0.0 } else { -0.0 }
+        }
+    } else {
+        // Normal
+        let new_exp = exp as i64 - 16383 + 1023;
+        if new_exp >= 2047 {
+            // Overflow to infinity
+            if sign == 0 {
+                f64::INFINITY
+            } else {
+                f64::NEG_INFINITY
+            }
+        } else if new_exp <= 0 {
+            // Underflow to zero
+            if sign == 0 { 0.0 } else { -0.0 }
+        } else {
+            // Normal f64
+            // Mantissa: take top 52 bits of 112-bit mantissa
+            let new_mant = (mant >> (112 - 52)) as u64;
+            let bits = (sign << 63) | ((new_exp as u64) << 52) | new_mant;
+            f64::from_bits(bits)
+        }
+    }
 }
 
 #[cfg(test)]

src/uu/od/src/input_decoder.rs

@@ -87,7 +87,13 @@ impl MultifileReader<'_> {
                             // print an error at the time that the file is needed,
                             // then move to the next file.
                             // This matches the behavior of the original `od`
-                            show_error!("{}: {e}", fname.maybe_quote());
+                            // Format error without OS error code to match GNU od
+                            let error_msg = match e.kind() {
+                                io::ErrorKind::NotFound => "No such file or directory",
+                                io::ErrorKind::PermissionDenied => "Permission denied",
+                                _ => "I/O error",
+                            };
+                            show_error!("{}: {}", fname.maybe_quote().external(true), error_msg);
                             self.any_err = true;
                         }
                     }

src/uu/od/src/multifile_reader.rs

@@ -80,14 +80,19 @@ struct OdOptions {
 }
 
 /// Helper function to parse bytes with error handling
-fn parse_bytes_option(matches: &ArgMatches, option_name: &str) -> UResult<Option<u64>> {
+fn parse_bytes_option(
+    matches: &ArgMatches,
+    args: &[String],
+    option_name: &str,
+    short: Option<char>,
+) -> UResult<Option<u64>> {
     match matches.get_one::<String>(option_name) {
         None => Ok(None),
         Some(s) => match parse_number_of_bytes(s) {
             Ok(n) => Ok(Some(n)),
             Err(e) => Err(USimpleError::new(
                 1,
-                format_error_message(&e, s, option_name),
+                format_error_message(&e, s, &option_display_name(args, option_name, short)),
             )),
         },
     }
@@ -110,12 +115,12 @@ impl OdOptions {
             ByteOrder::Native
         };
 
-        let mut skip_bytes = parse_bytes_option(matches, options::SKIP_BYTES)?.unwrap_or(0);
+        let mut skip_bytes =
+            parse_bytes_option(matches, args, options::SKIP_BYTES, Some('j'))?.unwrap_or(0);
 
         let mut label: Option<u64> = None;
 
-        let parsed_input = parse_inputs(matches)
-            .map_err(|e| USimpleError::new(1, translate!("od-error-invalid-inputs", "msg" => e)))?;
+        let parsed_input = parse_inputs(matches).map_err(|e| USimpleError::new(1, e))?;
         let input_strings = match parsed_input {
             CommandLineInputs::FileNames(v) => v,
             CommandLineInputs::FileAndOffset((f, s, l)) => {
@@ -131,16 +136,30 @@ impl OdOptions {
             None => 16,
             Some(s) => {
                 if matches.value_source(options::WIDTH) == Some(ValueSource::CommandLine) {
-                    match parse_number_of_bytes(s) {
-                        Ok(n) => usize::try_from(n)
-                            .map_err(|_| USimpleError::new(1, format!("‘{s}‘ is too large")))?,
-                        Err(e) => {
-                            return Err(USimpleError::new(
-                                1,
-                                format_error_message(&e, s, options::WIDTH),
-                            ));
-                        }
+                    let width_display = option_display_name(args, options::WIDTH, Some('w'));
+                    let parsed = parse_number_of_bytes(s).map_err(|e| {
+                        USimpleError::new(1, format_error_message(&e, s, &width_display))
+                    })?;
+                    if parsed == 0 {
+                        return Err(USimpleError::new(
+                            1,
+                            translate!(
+                                "od-error-invalid-argument",
+                                "option" => width_display.clone(),
+                                "value" => s.quote()
+                            ),
+                        ));
                     }
+                    usize::try_from(parsed).map_err(|_| {
+                        USimpleError::new(
+                            1,
+                            translate!(
+                                "od-error-argument-too-large",
+                                "option" => width_display.clone(),
+                                "value" => s.quote()
+                            ),
+                        )
+                    })?
                 } else {
                     16
                 }
@@ -160,9 +179,9 @@ impl OdOptions {
 
         let output_duplicates = matches.get_flag(options::OUTPUT_DUPLICATES);
 
-        let read_bytes = parse_bytes_option(matches, options::READ_BYTES)?;
+        let read_bytes = parse_bytes_option(matches, args, options::READ_BYTES, Some('N'))?;
 
-        let string_min_length = match parse_bytes_option(matches, options::STRINGS)? {
+        let string_min_length = match parse_bytes_option(matches, args, options::STRINGS, Some('S'))? {
             None => None,
             Some(n) => Some(usize::try_from(n).map_err(|_| {
                 USimpleError::new(
@@ -491,7 +510,9 @@ where
                 let length = memory_decoder.length();
 
                 if length == 0 {
-                    input_offset.print_final_offset();
+                    if !input_decoder.has_error() {
+                        input_offset.print_final_offset();
+                    }
                     break;
                 }
 
@@ -669,6 +690,10 @@ fn print_bytes(prefix: &str, input_decoder: &MemoryDecoder, output_info: &Output
                     let p = input_decoder.read_float(b, f.formatter_item_info.byte_size);
                     output_text.push_str(&func(p));
                 }
+                FormatWriter::LongDoubleWriter(func) => {
+                    let p = input_decoder.read_long_double(b);
+                    output_text.push_str(&func(p));
+                }
                 FormatWriter::BFloatWriter(func) => {
                     let p = input_decoder.read_bfloat(b);
                     output_text.push_str(&func(p));
@@ -745,6 +770,27 @@ impl<R: HasError> HasError for BufReader<R> {
     }
 }
 
+fn option_display_name(args: &[String], option_name: &str, short: Option<char>) -> String {
+    let long_form = format!("--{option_name}");
+    let long_form_with_eq = format!("{long_form}=");
+    if let Some(short_char) = short {
+        let short_form = format!("-{short_char}");
+        for arg in args.iter().skip(1) {
+            if !arg.starts_with("--") && arg.starts_with(&short_form) {
+                return short_form;
+            }
+        }
+        for arg in args.iter().skip(1) {
+            if arg == &long_form || arg.starts_with(&long_form_with_eq) {
+                return long_form;
+            }
+        }
+        short_form
+    } else {
+        long_form
+    }
+}
+
 fn format_error_message(error: &ParseSizeError, s: &str, option: &str) -> String {
     // NOTE:
     // GNU's od echos affected flag, -N or --read-bytes (-j or --skip-bytes, etc.), depending user's selection