Author: David Spickett Date: 2026-02-26T09:42:43Z New Revision: 0d8b3362f821c1b6ee11668c64441bf3a4b6ecdc
URL: https://github.com/llvm/llvm-project/commit/0d8b3362f821c1b6ee11668c64441bf3a4b6ecdc DIFF: https://github.com/llvm/llvm-project/commit/0d8b3362f821c1b6ee11668c64441bf3a4b6ecdc.diff LOG: [lldb] Fix partial Unicode handling in TrimAndPad (#183299) This fixes an issue I found while writing a variant of TrimAndPad for word wrapping purposes. TrimAndPad relies on llvm::sys::locale::ColumnWidth returning an error value, when there is a partial uft8 character. This error code is ErrorInvalidUTF8, which is -2. The idea was that when you cast that into a size_t, it was always going to be larger than our target visible width. The check for this was: result_visibile_length + column_width <= visible_length Where result_visible_length and column_width are size_ts. And the idea is correct, as long as result_visible_width is not large enough to cause the result to wrap to be lower than visible_length. In other words: when we have accumulated some part of the string already, if we have to trim the next part, sometimes we would return an incomplete Unicode character. When we should instead carry on trimming. To fix this, I've made TrimAndPad check explicitly for the error code. Only if it's not found, will it cast to do the second check. Added: Modified: lldb/include/lldb/Utility/AnsiTerminal.h lldb/unittests/Utility/AnsiTerminalTest.cpp Removed: ################################################################################ diff --git a/lldb/include/lldb/Utility/AnsiTerminal.h b/lldb/include/lldb/Utility/AnsiTerminal.h index 153602cc08b09..dd2eafef1f228 100644 --- a/lldb/include/lldb/Utility/AnsiTerminal.h +++ b/lldb/include/lldb/Utility/AnsiTerminal.h @@ -98,6 +98,7 @@ #include "llvm/ADT/STLExtras.h" #include "llvm/ADT/StringRef.h" #include "llvm/Support/Locale.h" +#include "llvm/Support/Unicode.h" #include "lldb/Utility/Stream.h" @@ -264,13 +265,16 @@ inline std::string TrimAndPad(llvm::StringRef str, size_t visible_length, // Repeatedly trim the string until it its valid unicode and fits. llvm::StringRef trimmed = left; while (!trimmed.empty()) { - // This relies on columnWidth returning -2 for invalid/partial unicode - // characters, which after conversion to size_t will be larger than the - // visible width. - column_width = llvm::sys::locale::columnWidth(trimmed); - if (result_visibile_length + column_width <= visible_length) { + int trimmed_width = llvm::sys::locale::columnWidth(trimmed); + if ( + // If we have only part of a Unicode character, keep trimming. + trimmed_width != + llvm::sys::unicode::ColumnWidthErrors::ErrorInvalidUTF8 && + // If the trimmed string fits, take it. + result_visibile_length + static_cast<size_t>(trimmed_width) <= + visible_length) { result.append(trimmed); - result_visibile_length += column_width; + result_visibile_length += static_cast<size_t>(trimmed_width); break; } trimmed = trimmed.drop_back(); diff --git a/lldb/unittests/Utility/AnsiTerminalTest.cpp b/lldb/unittests/Utility/AnsiTerminalTest.cpp index 28fa32461ad5f..8924c083a656b 100644 --- a/lldb/unittests/Utility/AnsiTerminalTest.cpp +++ b/lldb/unittests/Utility/AnsiTerminalTest.cpp @@ -116,6 +116,18 @@ TEST(AnsiTerminal, TrimAndPad) { EXPECT_EQ(" ❤️", ansi::TrimAndPad(" ❤️", 5)); EXPECT_EQ("12❤️4❤️", ansi::TrimAndPad("12❤️4❤️", 5)); EXPECT_EQ("12❤️45", ansi::TrimAndPad("12❤️45❤️", 5)); + + // This string previously triggered a bug in handling incomplete Unicode + // characters, when we had already accumulated some previous parts of the + // string. + const char *quick = "The \x1B[0mquick\x1B[0m 💨\x1B[0m"; + EXPECT_EQ(ansi::TrimAndPad(quick, 0), ""); + EXPECT_EQ(ansi::TrimAndPad(quick, 9), "The \x1B[0mquick\x1B[0m"); + EXPECT_EQ(ansi::TrimAndPad(quick, 10), "The \x1B[0mquick\x1B[0m "); + // The emoji is 2 columns, so 11 is not quite enough. + EXPECT_EQ(ansi::TrimAndPad(quick, 11, '_'), "The \x1B[0mquick\x1B[0m _"); + // 12 exactly enough to include the emoji and proceeding ANSI code. + EXPECT_EQ(ansi::TrimAndPad(quick, 12), quick); } static void TestLines(const std::string &input, int indent, _______________________________________________ lldb-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
