amuled -o corrupts non-ASCII log lines on stdout (uses C-locale wxConvLibc, never calls setlocale); on-disk logfile is fine #40

Summary

amuled -o (log to stdout) corrupts every log line that contains non-ASCII
characters — e.g. No shareable files found in directory: <path> for a shared
directory whose name has accented/Latin-1 bytes. The bytes come out as ?,
U+FFFD (EF BF BD) and stray control bytes, often losing the line's
timestamp/prefix and bleeding into adjacent lines.

The on-disk logfile (~/.aMule/logfile) for the very same lines is perfectly
fine (correct UTF-8). Only the stdout stream is corrupted.

amuled does not crash; the share scan completes normally.

Environment

• aMule 3.0.0 (CMake build), wxBase(GTK3) 3.2.8, Boost 1.83, Debian, x86-64
• amuled -o, headless, in a C/POSIX locale (e.g. a minimal container: LANG
  unset, LC_CTYPE=POSIX)

Minimal deterministic reproduction

mkdir -p /share/AAA_before $(printf '/share/bad_\x8e\x7f_x') $(printf '/share/caf\xe9_\xf1_y') /share/ZZZ_after
# share /share recursively, then run amuled in a C locale logging to stdout:
LC_ALL=C amuled -o -c <configdir>

stdout shows the two ASCII dirs cleanly but the two non-ASCII dirs come out as
mangled bytes with no timestamp/prefix. ~/.aMule/logfile shows all four lines
correctly (UTF-8).

Root cause

CLogger::DoLine writes each line to two sinks with different conversions
(src/Logger.cpp):

// logfile  -> UTF-8 (clean)
m_ApplogBuf += line;
FlushApplog();                 // wxStringInputStream => wxConvUTF8  (Logger.cpp:298-304)

// stdout   -> C-locale (corrupted)
if (m_StdoutLog || toStdout) {
    printf("%s", (const char*)unicode2char(line));   // Logger.cpp:274
}

unicode2char() converts via wxConvLibc — the C-library locale converter
(src/libs/common/StringFunctions.cpp:42-60). It only works if the current C
locale is UTF-8-capable.

But amuled never calls setlocale(LC_ALL, "") — that call exists only for the
external connectors (src/ExternalConnector.cpp:675, used by amulecmd/amuleweb),
not for CamuleDaemonApp. So amuled's C locale stays at the default "C"
regardless of LANG/LC_ALL, and wxConvLibc mangles every non-ASCII character
written to stdout. The logfile path uses wxConvUTF8 and is therefore unaffected.

Because the converter often emits incomplete/replacement sequences mid-buffer, the
mangled output also loses line boundaries and merges into neighbouring lines.

Evidence

• strace -f -e write shows amuled's actual write() to the logfile fd is clean,
  correctly-UTF-8-encoded, one full line per write.
• ~/.aMule/logfile on disk: clean UTF-8 for the exact same directories.
• The corrupted bytes are only ever on the stdout stream.
• Setting LANG/LC_ALL/LC_CTYPE=C.UTF-8 (or C.utf8) on the process has no
  effect — consistent with amuled never calling setlocale.

Suggested fix (either)

1. Make the stdout sink use UTF-8 to match the logfile, e.g. write
   line.utf8_str() instead of unicode2char(line) in CLogger::DoLine
   (Logger.cpp:274); or
2. Call setlocale(LC_ALL, "") during CamuleDaemonApp startup so wxConvLibc
   honours the environment locale (same as the connectors already do).

Option 1 is more robust since it does not depend on the deployment locale.