This is mostly a problem within avocado as serial generally isn't busy
enough to overfill pipes. However the consequences of recording a
failed write will haunt us on replay when the log will be out of sync
to the playback.

Fixes: https://gitlab.com/qemu-project/qemu/-/issues/2010
Acked-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20231211091346.14616-13-alex.bennee@linaro.org>

diff --git a/chardev/char.c b/chardev/char.c
index 996a024c7a2..48f28881c28 100644
--- a/chardev/char.c
+++ b/chardev/char.c
@@ -171,6 +171,18 @@ int qemu_chr_write(Chardev *s, const uint8_t *buf, int len, bool write_all)
         return res;
     }
 
+    if (replay_mode == REPLAY_MODE_RECORD) {
+        /*
+         * When recording we don't want temporary conditions to
+         * perturb the result. By ensuring we write everything we can
+         * while recording we avoid playback being out of sync if it
+         * doesn't encounter the same temporary conditions (usually
+         * triggered by external programs not reading the chardev fast
+         * enough and pipes filling up).
+         */
+        write_all = true;
+    }
+
     res = qemu_chr_write_buffer(s, buf, len, &offset, write_all);
 
     if (qemu_chr_replay(s) && replay_mode == REPLAY_MODE_RECORD) {
-- 
2.39.2