From: Philippe Proulx <eeppeliteloop@gmail.com>
Date: Fri, 10 May 2024 20:12:52 +0000 (-0400)
Subject: src.ctf.*: add `ctf::src::NullCpFinder`
X-Git-Url: http://drtracing.org/?a=commitdiff_plain;h=cd34c1643b882fafbee70f9600a884351eb55486;p=babeltrace.git

src.ctf.*: add `ctf::src::NullCpFinder`

An instance of this new class keeps a temporary code unit of length
`CodeUnitLenV` bytes to help finder a null codepoint (U+0000) in
contiguous buffers (especially when the encoded codepoint spans more
than one buffer).

See the comment of the new class to learn more.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I80feb4402cc0256d7c48c357ca23e355c4322ce0
Reviewed-on: https://review.lttng.org/c/babeltrace/+/12707
---

diff --git a/src/Makefile.am b/src/Makefile.am
index 24efd77e..2c92803c 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -691,6 +691,7 @@ plugins_ctf_babeltrace_plugin_ctf_la_SOURCES = \
 	plugins/ctf/common/src/metadata/ctf-ir.hpp \
 	plugins/ctf/common/src/msg-iter/msg-iter.cpp \
 	plugins/ctf/common/src/msg-iter/msg-iter.hpp \
+	plugins/ctf/common/src/null-cp-finder.hpp \
 	plugins/ctf/fs-sink/fs-sink.cpp \
 	plugins/ctf/fs-sink/fs-sink-ctf-meta.hpp \
 	plugins/ctf/fs-sink/fs-sink.hpp \
diff --git a/src/plugins/ctf/common/src/null-cp-finder.hpp b/src/plugins/ctf/common/src/null-cp-finder.hpp
new file mode 100644
index 00000000..e804c4e1
--- /dev/null
+++ b/src/plugins/ctf/common/src/null-cp-finder.hpp
@@ -0,0 +1,125 @@
+/*
+ * Copyright (c) 2024 Philippe Proulx <pproulx@efficios.com>
+ *
+ * SPDX-License-Identifier: MIT
+ */
+
+#ifndef BABELTRACE_PLUGINS_CTF_COMMON_SRC_NULL_CP_FINDER_HPP
+#define BABELTRACE_PLUGINS_CTF_COMMON_SRC_NULL_CP_FINDER_HPP
+
+#include <array>
+#include <cstdlib>
+
+#include "cpp-common/bt2c/aliases.hpp"
+#include "cpp-common/bt2s/optional.hpp"
+
+namespace ctf {
+namespace src {
+
+/*
+ * Null (U+0000) codepoint finder.
+ *
+ * An instance of this class keeps a temporary code unit of length
+ * `CodeUnitLenV` bytes.
+ *
+ * Call findNullCp() to try to find the first U+0000 codepoint within
+ * some string data part, updating the state of the finder at the same
+ * time.
+ *
+ * When decoding a UTF-16/UTF-32 null-terminated string, the bytes of
+ * the encoded U+0000 codepoint may span more than one medium buffer.
+ *
+ * For example, consider this scenario:
+ *
+ *     ╔═══════════════════════════════════════════════════════════════╗
+ *     ║                            This null byte part of             ║
+ *     ║                            the first U+0000 codepoint         ║
+ *     ║                            ┆                                  ║
+ *     ║ Current buffer             ┆     Next buffer                  ║
+ *     ║ ┈┈┈────────────────────────▼──┐ ┌────────────────────────┈┈┈  ║
+ *     ║    64 00 20 00 3c d8 3b df 00 │ │ 00 1f fc cc bc 44 35 56     ║
+ *     ║ ┈┈┈▲──────────────────────────┘ └─▲──────────────────────┈┈┈  ║
+ *     ║    ┆                              ┆                           ║
+ *     ║    Code unit                      This null byte also part of ║
+ *     ║    beginning                      the first U+0000 codepoint  ║
+ *     ╚═══════════════════════════════════════════════════════════════╝
+ *
+ * Assume a UTF-16LE encoding (code unit size is two). Then there are
+ * four complete code units in the current buffer, and half of one (the
+ * last null byte).
+ *
+ * The two null bytes of the first U+0000 codepoint are within two
+ * different buffers.
+ *
+ * The strategy here is keep each code unit in a temporary buffer
+ * (`_mCodeUnitBuf`), along with its length in bytes
+ * (`_mCodeUnitBufLen`). In findNullCp(), when `_mCodeUnitBufLen`
+ * reaches the `CodeUnitLenV` while decoding, then we check if it
+ * encodes U+0000.
+ *
+ * In the example above, after reading that the null byte of the current
+ * buffer, `_mCodeUnitBuf[0]` is zero and `_mCodeUnitBufLen` is one.
+ *
+ * Afterwards, when given the next buffer, findNullCp() continues
+ * reading the current code unit, making `_mCodeUnitBuf[1]` zero and
+ * `_mCodeUnitBufLen` two. Since `_mCodeUnitBufLen` is equal to the code
+ * unit size, the method can check the current code unit value: two
+ * zeros, which means U+0000, which means the end of that
+ * null-terminated string.
+ */
+template <std::size_t CodeUnitLenV>
+class NullCpFinder final
+{
+    static_assert(CodeUnitLenV == 1 || CodeUnitLenV == 2 || CodeUnitLenV == 4,
+                  "`CodeUnitLenV` is 1 (UTF-8), 2 (UTF-16), or 4 (UTF-32).");
+
+public:
+    explicit NullCpFinder() = default;
+
+    /*
+     * Tries to find the first U+0000 codepoint in `buffer` considering
+     * what you already passed to this method and `CodeUnitLenV`.
+     *
+     * Returns an iterator _after_ the end of the encoded U+0000
+     * codepoint on success, or `bt2s::nullopt` when it didn't find any
+     * U+0000 codepoint. This means this method may return `str.end()`
+     * if `str` finishes with a U+0000 codepoint.
+     */
+    bt2s::optional<bt2c::ConstBytes::const_iterator>
+    findNullCp(const bt2c::ConstBytes buffer) noexcept
+    {
+        for (auto it = buffer.begin(); it != buffer.end(); ++it) {
+            _mCodeUnitBuf[_mCodeUnitBufLen] = *it;
+            ++_mCodeUnitBufLen;
+
+            if (_mCodeUnitBufLen == CodeUnitLenV) {
+                /* New complete code unit: is it U+0000? */
+                if (_mCodeUnitBuf == _CodeUnitBuf {0}) {
+                    /* Found U+0000 */
+                    return it + 1;
+                }
+
+                /* New empty code unit */
+                _mCodeUnitBufLen = 0;
+            }
+        }
+
+        /* No U+0000 codepoint found */
+        return bt2s::nullopt;
+    }
+
+private:
+    /* Code unit buffer type */
+    using _CodeUnitBuf = std::array<char, CodeUnitLenV>;
+
+    /* Code unit buffer */
+    _CodeUnitBuf _mCodeUnitBuf;
+
+    /* Code unit buffer length */
+    std::size_t _mCodeUnitBufLen = 0;
+};
+
+} /* namespace src */
+} /* namespace ctf */
+
+#endif /* BABELTRACE_PLUGINS_CTF_COMMON_SRC_NULL_CP_FINDER_HPP */