StringRef no longer repoints m_start to m_data after c_str() on a substring.

This fixes an issue where a self-assignment of a StringRef copy would point into internally (and now dangling) data.
(now self-assignment check is no longer needed)
This commit is contained in:
Phil Nash 2019-05-21 00:04:19 +01:00
parent 71fce429af
commit 96355da34e
3 changed files with 10 additions and 11 deletions

View file

@ -65,7 +65,6 @@ TEST_CASE( "StringRef", "[Strings][StringRef]" ) {
original.c_str(); // Forces it to take ownership
REQUIRE( isSubstring( original ) == false );
REQUIRE( isOwned( original ) );
}
@ -88,14 +87,13 @@ TEST_CASE( "StringRef", "[Strings][StringRef]" ) {
REQUIRE( rawChars == s.currentData() ); // same pointer value
REQUIRE( ss.c_str() != rawChars );
REQUIRE( isSubstring( ss ) == false );
REQUIRE( isOwned( ss ) );
REQUIRE( ss.currentData() != s.currentData() ); // different pointer value
SECTION( "Self-assignment after substring" ) {
ss = *&ss; // the *& are there to suppress warnings (see: "Improvements to Clang's diagnostics" in https://rev.ng/gitlab/revng-bar-2019/clang/raw/master/docs/ReleaseNotes.rst)
REQUIRE(isOwned(ss) == true);
REQUIRE( isOwned(ss) == false );
REQUIRE( ss == "hello" );
REQUIRE( rawChars == ss.currentData() ); // same pointer value
}
}