fix(scriptanalysis): M3.6 — full-framework analysis refs close forbidden-type-in-allowed-ns blind spot; pin Process/Stopwatch; fix stale codec test; drop dead ContainsInCode

This commit is contained in:
Joseph Doherty
2026-06-16 20:00:28 -04:00
parent cf935d5744
commit 069757209a
6 changed files with 121 additions and 127 deletions
@@ -2,12 +2,11 @@ namespace ZB.MOM.WW.ScadaBridge.TemplateEngine.Validation;
/// <summary>
/// String/comment-aware scanner for the balanced-delimiter ("does it look like
/// valid C#") checks used by <see cref="ScriptCompiler"/> and
/// <c>SharedScriptService.ValidateSyntax</c>.
/// valid C#") check used by <c>SharedScriptService.ValidateSyntax</c>.
///
/// <para>
/// This is <b>not</b> a compiler. It is an interim structural check that walks
/// the source once and tracks <c>{}</c>, <c>[]</c> and <c>()</c> depth while
/// This is <b>not</b> a compiler. It is a structural check that walks the
/// source once and tracks <c>{}</c>, <c>[]</c> and <c>()</c> depth while
/// correctly skipping over the C# lexical constructs in which a delimiter is
/// inert: line/block comments, regular string literals (with <c>\</c> escapes),
/// verbatim strings (<c>@"..."</c>, where <c>""</c> escapes a quote and <c>\</c>
@@ -17,11 +16,11 @@ namespace ZB.MOM.WW.ScadaBridge.TemplateEngine.Validation;
/// </para>
///
/// <para>
/// It is intentionally conservative: when the real Roslyn-based compiler is
/// wired in (see <see cref="ScriptCompiler"/>) this hand-rolled scan should be
/// replaced by <c>CSharpSyntaxTree.ParseText</c> diagnostics. Until then this
/// scanner removes the false positives that a naive character count produced
/// for valid scripts containing a delimiter inside a string or comment.
/// Trust enforcement and full compilation are NOT done here — those are the
/// authoritative <see cref="ScriptCompiler"/> (which delegates to the shared
/// <c>ZB.MOM.WW.ScadaBridge.ScriptAnalysis</c> validator + Roslyn compile). This
/// scanner only provides <c>SharedScriptService</c> a cheap pre-compile sanity
/// check for balanced delimiters.
/// </para>
/// </summary>
internal static class CSharpDelimiterScanner
@@ -41,121 +40,6 @@ internal static class CSharpDelimiterScanner
UnterminatedChar,
}
/// <summary>
/// Returns true when <paramref name="pattern"/> occurs in a <b>code</b>
/// region of <paramref name="code"/> — i.e. not wholly inside a string
/// literal, char literal, or comment. Used by the interim forbidden-API
/// scan so that the inert text <c>System.IO.</c> in a comment or string
/// literal is not flagged as a forbidden API call (TemplateEngine-006).
///
/// <para>
/// This removes the false-positive half of the substring scan. It does
/// <b>not</b> close the bypass half: namespace aliases, <c>using static</c>,
/// and <c>global::</c>-qualified references still evade a pure text match.
/// Authoritative forbidden-API enforcement requires Roslyn semantic symbol
/// analysis and is deferred to the real script compiler / Site Runtime
/// sandbox; this check is advisory only.
/// </para>
/// </summary>
/// <param name="code">The C# source code to scan.</param>
/// <param name="pattern">The substring to search for in code regions only.</param>
/// <returns><c>true</c> if <paramref name="pattern"/> occurs in a code region (not inside a comment, string, or char literal); otherwise <c>false</c>.</returns>
internal static bool ContainsInCode(string code, string pattern)
{
if (string.IsNullOrEmpty(pattern))
return false;
// Blank out every string/char-literal/comment span, then do an ordinary
// substring search over what remains (the code regions).
var codeOnly = BlankNonCodeSpans(code);
return codeOnly.Contains(pattern, StringComparison.Ordinal);
}
/// <summary>
/// Replaces the content of every comment, string literal, and char literal
/// with spaces (newlines preserved), leaving only code regions intact.
/// Delimiter characters themselves are also blanked so a pattern cannot
/// straddle a literal boundary.
/// </summary>
private static string BlankNonCodeSpans(string code)
{
var buffer = code.ToCharArray();
int n = code.Length;
int i = 0;
void Blank(int from, int to)
{
for (int k = from; k < to && k < n; k++)
if (buffer[k] != '\n' && buffer[k] != '\r')
buffer[k] = ' ';
}
while (i < n)
{
char c = code[i];
char next = i + 1 < n ? code[i + 1] : '\0';
int start = i;
if (c == '/' && next == '/')
{
i += 2;
while (i < n && code[i] != '\n') i++;
Blank(start, i);
continue;
}
if (c == '/' && next == '*')
{
i += 2;
while (i < n && !(code[i] == '*' && i + 1 < n && code[i + 1] == '/')) i++;
if (i < n) i += 2;
Blank(start, i);
continue;
}
if (c == '"' && next == '"' && i + 2 < n && code[i + 2] == '"')
{
SkipRawString(code, ref i);
Blank(start, i);
continue;
}
if (c == '$')
{
int j = i + 1;
bool verbatim = false;
if (j < n && code[j] == '@') { verbatim = true; j++; }
if (j < n && code[j] == '"')
{
i = j;
SkipInterpolatedString(code, ref i, verbatim);
Blank(start, i);
continue;
}
}
if (c == '@' && next == '"')
{
i++;
SkipVerbatimString(code, ref i);
Blank(start, i);
continue;
}
if (c == '"')
{
SkipRegularString(code, ref i);
Blank(start, i);
continue;
}
if (c == '\'')
{
SkipCharLiteral(code, ref i);
Blank(start, i);
continue;
}
i++;
}
return new string(buffer);
}
/// <summary>
/// Walks <paramref name="code"/> once and reports the first structural
/// delimiter problem, or <see cref="Mismatch.None"/> when the source is