[flang] Enforce fixed form rules about END continuation
From subclause 6.3.3.5: a program unit END statement cannot be
continued in fixed form, and other statements cannot have initial
lines that look like program unit END statements. I think this
is to avoid violating assumptions that are important to legacy
compilers' statement classification routines.
Differential Revision: https://reviews.llvm.org/D109933
GitOrigin-RevId: f6ddfac401b8a5dfbf7623ebd8800df5184fa244
diff --git a/lib/Parser/prescan.cpp b/lib/Parser/prescan.cpp
index 18d69d1..1408fdf 100644
--- a/lib/Parser/prescan.cpp
+++ b/lib/Parser/prescan.cpp
@@ -217,6 +217,9 @@
if (line.kind == LineClassification::Kind::CompilerDirective) {
SourceFormChange(tokens.ToString());
}
+ if (inFixedForm_ && line.kind == LineClassification::Kind::Source) {
+ EnforceStupidEndStatementRules(tokens);
+ }
tokens.CheckBadFortranCharacters(messages_).Emit(cooked_);
}
if (omitNewline_) {
@@ -288,6 +291,67 @@
}
}
+// 6.3.3.5: A program unit END statement, or any other statement whose
+// initial line resembles an END statement, shall not be continued in
+// fixed form source.
+void Prescanner::EnforceStupidEndStatementRules(const TokenSequence &tokens) {
+ CharBlock cBlock{tokens.ToCharBlock()};
+ const char *str{cBlock.begin()};
+ std::size_t n{cBlock.size()};
+ if (n < 3) {
+ return;
+ }
+ std::size_t j{0};
+ for (; j < n && (str[j] == ' ' || (str[j] >= '0' && str[j] <= '9')); ++j) {
+ }
+ if (j + 3 > n || std::memcmp(str + j, "end", 3) != 0) {
+ return;
+ }
+ // It starts with END, possibly after a label.
+ auto start{allSources_.GetSourcePosition(tokens.GetCharProvenance(j))};
+ auto end{allSources_.GetSourcePosition(tokens.GetCharProvenance(n - 1))};
+ if (!start || !end) {
+ return;
+ }
+ if (&start->file == &end->file && start->line == end->line) {
+ return; // no continuation
+ }
+ j += 3;
+ static const char *const prefixes[]{"program", "subroutine", "function",
+ "blockdata", "module", "submodule", nullptr};
+ CharBlock stmt{tokens.ToCharBlock()};
+ bool isPrefix{j == n || !IsLegalInIdentifier(str[j])}; // prefix is END
+ std::size_t endOfPrefix{j - 1};
+ for (const char *const *p{prefixes}; *p; ++p) {
+ std::size_t pLen{std::strlen(*p)};
+ if (j + pLen <= n && std::memcmp(str + j, *p, pLen) == 0) {
+ isPrefix = true; // END thing as prefix
+ j += pLen;
+ endOfPrefix = j - 1;
+ for (; j < n && IsLegalInIdentifier(str[j]); ++j) {
+ }
+ break;
+ }
+ }
+ if (isPrefix) {
+ auto range{tokens.GetTokenProvenanceRange(1)};
+ if (j == n) { // END or END thing [name]
+ Say(range,
+ "Program unit END statement may not be continued in fixed form source"_err_en_US);
+ } else {
+ auto endOfPrefixPos{
+ allSources_.GetSourcePosition(tokens.GetCharProvenance(endOfPrefix))};
+ auto next{allSources_.GetSourcePosition(tokens.GetCharProvenance(j))};
+ if (endOfPrefixPos && next && &endOfPrefixPos->file == &start->file &&
+ endOfPrefixPos->line == start->line &&
+ (&next->file != &start->file || next->line != start->line)) {
+ Say(range,
+ "Initial line of continued statement must not appear to be a program unit END in fixed form source"_err_en_US);
+ }
+ }
+ }
+}
+
void Prescanner::SkipToEndOfLine() {
while (*at_ != '\n') {
++at_, ++column_;
diff --git a/lib/Parser/prescan.h b/lib/Parser/prescan.h
index 2f1b83e..4d002ad 100644
--- a/lib/Parser/prescan.h
+++ b/lib/Parser/prescan.h
@@ -150,6 +150,7 @@
}
void LabelField(TokenSequence &);
+ void EnforceStupidEndStatementRules(const TokenSequence &);
void SkipToEndOfLine();
bool MustSkipToEndOfLine() const;
void NextChar();
diff --git a/lib/Parser/token-sequence.cpp b/lib/Parser/token-sequence.cpp
index 0a959f2..3ed9d05 100644
--- a/lib/Parser/token-sequence.cpp
+++ b/lib/Parser/token-sequence.cpp
@@ -286,10 +286,14 @@
return o;
}
+Provenance TokenSequence::GetCharProvenance(std::size_t offset) const {
+ ProvenanceRange range{provenances_.Map(offset)};
+ return range.start();
+}
+
Provenance TokenSequence::GetTokenProvenance(
std::size_t token, std::size_t offset) const {
- ProvenanceRange range{provenances_.Map(start_[token] + offset)};
- return range.start();
+ return GetCharProvenance(start_[token] + offset);
}
ProvenanceRange TokenSequence::GetTokenProvenanceRange(
diff --git a/lib/Parser/token-sequence.h b/lib/Parser/token-sequence.h
index 16cef37..6da6229 100644
--- a/lib/Parser/token-sequence.h
+++ b/lib/Parser/token-sequence.h
@@ -102,6 +102,7 @@
void Put(const std::string &, Provenance);
void Put(llvm::raw_string_ostream &, Provenance);
+ Provenance GetCharProvenance(std::size_t) const;
Provenance GetTokenProvenance(
std::size_t token, std::size_t offset = 0) const;
ProvenanceRange GetTokenProvenanceRange(
diff --git a/test/Parser/end.f b/test/Parser/end.f
new file mode 100644
index 0000000..e0225e6
--- /dev/null
+++ b/test/Parser/end.f
@@ -0,0 +1,29 @@
+! RUN: not %flang_fc1 -fsyntax-only %s 2>&1 | FileCheck %s
+! CHECK: end.f:3:7: error: Program unit END statement may not be continued in fixed form source
+ e
+ + nd
+! CHECK: end.f:6:7: error: Program unit END statement may not be continued in fixed form source
+ end prog
+ + ram
+! CHECK: end.f:9:7: error: Program unit END statement may not be continued in fixed form source
+ end
+ + program
+! CHECK: end.f:12:7: error: Program unit END statement may not be continued in fixed form source
+ end
+ + program
+ 1 main
+! CHECK: end.f:16:7: error: Program unit END statement may not be continued in fixed form source
+ end program
+ 1 main
+! CHECK: end.f:19:7: error: Initial line of continued statement must not appear to be a program unit END in fixed form source
+ end
+ + = end + 1
+! CHECK: end.f:22:7: error: Initial line of continued statement must not appear to be a program unit END in fixed form source
+ end module
+ + = end module + 1
+! CHECK-NOT: end.f:25:7: error: Initial line of continued statement must not appear to be a program unit END in fixed form source
+ end =
+ + end + 1
+! CHECK-NOT: end.f:28:7: error: Initial line of continued statement must not appear to be a program unit END in fixed form source
+ end block data (
+ + 1) = 666
