Detect NulInCStr error earlier.

By making it an `EscapeError` instead of a `LitError`. This makes it
more like the other errors produced during unescaping.

NOTE: this means these errors are issued earlier, before expansion,
which changes behaviour. The next commit will delay issue of this error
and others, reverting the behaviour change for this particular error.

One nice thing about this: the old approach had some code in
`report_lit_error` to calculate the span of the nul char from a range.
This code used a hardwired `+2` to account for the `c"` at the start of
a C string literal, but this should have changed to a `+3` for raw C
string literals to account for the `cr"`, which meant that the caret in
`cr"` nul error messages was one short of where it should have been. The
new approach doesn't need any of this and avoids the off-by-one error.

Author: nnethercote

compiler/rustc_ast/src/util/literal.rs

@@ -8,7 +8,6 @@ use rustc_lexer::unescape::{
 };
 use rustc_span::symbol::{kw, sym, Symbol};
 use rustc_span::Span;
-use std::ops::Range;
 use std::{ascii, fmt, str};
 
 // Escapes a string, represented as a symbol. Reuses the original symbol,
@@ -39,7 +38,6 @@ pub enum LitError {
     InvalidFloatSuffix,
     NonDecimalFloat(u32),
     IntTooLarge(u32),
-    NulInCStr(Range<usize>),
 }
 
 impl LitKind {
@@ -156,10 +154,7 @@ impl LitKind {
                 let s = symbol.as_str();
                 let mut buf = Vec::with_capacity(s.len());
                 let mut error = Ok(());
-                unescape_c_string(s, Mode::CStr, &mut |span, c| match c {
-                    Ok(CStrUnit::Byte(0) | CStrUnit::Char('\0')) => {
-                        error = Err(LitError::NulInCStr(span));
-                    }
+                unescape_c_string(s, Mode::CStr, &mut |_span, c| match c {
                     Ok(CStrUnit::Byte(b)) => buf.push(b),
                     Ok(CStrUnit::Char(c)) => {
                         buf.extend_from_slice(c.encode_utf8(&mut [0; 4]).as_bytes())
@@ -179,10 +174,7 @@ impl LitKind {
                 // can convert the symbol directly to a `Lrc<u8>` on success.
                 let s = symbol.as_str();
                 let mut error = Ok(());
-                unescape_c_string(s, Mode::RawCStr, &mut |span, c| match c {
-                    Ok(CStrUnit::Byte(0) | CStrUnit::Char('\0')) => {
-                        error = Err(LitError::NulInCStr(span));
-                    }
+                unescape_c_string(s, Mode::RawCStr, &mut |_, c| match c {
                     Ok(_) => {}
                     Err(err) => {
                         if err.is_fatal() {

compiler/rustc_lexer/src/unescape.rs

@@ -59,6 +59,9 @@ pub enum EscapeError {
     /// Non-ascii character in byte literal, byte string literal, or raw byte string literal.
     NonAsciiCharInByte,
 
+    // `\0` in a C string literal.
+    NulInCStr,
+
     /// After a line ending with '\', the next line contains whitespace
     /// characters that are not skipped.
     UnskippedWhitespaceWarning,
@@ -122,10 +125,18 @@ where
 {
     match mode {
         CStr => {
-            unescape_str_common(src, mode, callback);
+            unescape_str_common(src, mode, &mut |r, mut result| {
+                if let Ok(CStrUnit::Byte(0) | CStrUnit::Char('\0')) = result {
+                    result = Err(EscapeError::NulInCStr);
+                }
+                callback(r, result)
+            });
         }
         RawCStr => {
-            unescape_raw_str_or_raw_byte_str(src, mode, &mut |r, result| {
+            unescape_raw_str_or_raw_byte_str(src, mode, &mut |r, mut result| {
+                if let Ok('\0') = result {
+                    result = Err(EscapeError::NulInCStr);
+                }
                 callback(r, result.map(CStrUnit::Char))
             });
         }

compiler/rustc_parse/messages.ftl

@@ -612,6 +612,8 @@ parse_note_mut_pattern_usage = `mut` may be followed by `variable` and `variable
 
 parse_note_pattern_alternatives_use_single_vert = alternatives in or-patterns are separated with `|`, not `||`
 
+parse_nul_in_c_str = null characters in C string literals are not supported
+
 parse_or_pattern_not_allowed_in_fn_parameters = top-level or-patterns are not allowed in function parameters
 parse_or_pattern_not_allowed_in_let_binding = top-level or-patterns are not allowed in `let` bindings
 parse_out_of_range_hex_escape = out of range hex escape

compiler/rustc_parse/src/errors.rs

@@ -2138,6 +2138,11 @@ pub enum UnescapeError {
         #[subdiagnostic]
         suggestion: MoreThanOneCharSugg,
     },
+    #[diag(parse_nul_in_c_str)]
+    NulInCStr {
+        #[primary_span]
+        span: Span,
+    },
 }
 
 #[derive(Subdiagnostic)]

compiler/rustc_parse/src/lexer/unescape_error_reporting.rs

@@ -265,6 +265,9 @@ pub(crate) fn emit_unescape_error(
         EscapeError::LoneSlash => {
             handler.emit_err(UnescapeError::LoneSlash(err_span));
         }
+        EscapeError::NulInCStr => {
+            handler.emit_err(UnescapeError::NulInCStr { span: err_span });
+        }
         EscapeError::UnskippedWhitespaceWarning => {
             let (c, char_span) = last_char();
             handler.emit_warning(UnescapeError::UnskippedWhitespace {

compiler/rustc_session/messages.ftl

@@ -72,8 +72,6 @@ session_not_circumvent_feature = `-Zunleash-the-miri-inside-of-you` may not be u
 
 session_not_supported = not supported
 
-session_nul_in_c_str = null characters in C string literals are not supported
-
 session_octal_float_literal_not_supported = octal float literal is not supported
 
 session_optimization_fuel_exhausted = optimization-fuel-exhausted: {$msg}

compiler/rustc_session/src/errors.rs

@@ -5,7 +5,7 @@ use rustc_ast::token;
 use rustc_ast::util::literal::LitError;
 use rustc_errors::{error_code, DiagnosticMessage, ErrorGuaranteed, IntoDiagnostic, MultiSpan};
 use rustc_macros::Diagnostic;
-use rustc_span::{BytePos, Span, Symbol};
+use rustc_span::{Span, Symbol};
 use rustc_target::spec::{SplitDebuginfo, StackProtector, TargetTriple};
 
 pub struct FeatureGateError {
@@ -329,13 +329,6 @@ pub(crate) struct BinaryFloatLiteralNotSupported {
     pub span: Span,
 }
 
-#[derive(Diagnostic)]
-#[diag(session_nul_in_c_str)]
-pub(crate) struct NulInCStr {
-    #[primary_span]
-    pub span: Span,
-}
-
 pub fn report_lit_error(sess: &ParseSess, err: LitError, lit: token::Lit, span: Span) {
     // Checks if `s` looks like i32 or u1234 etc.
     fn looks_like_width_suffix(first_chars: &[char], s: &str) -> bool {
@@ -414,12 +407,6 @@ pub fn report_lit_error(sess: &ParseSess, err: LitError, lit: token::Lit, span:
             };
             sess.emit_err(IntLiteralTooLarge { span, limit });
         }
-        LitError::NulInCStr(range) => {
-            let lo = BytePos(span.lo().0 + range.start as u32 + 2);
-            let hi = BytePos(span.lo().0 + range.end as u32 + 2);
-            let span = span.with_lo(lo).with_hi(hi);
-            sess.emit_err(NulInCStr { span });
-        }
     }
 }