gdb_server: fix a comment in gdb_new_connection()
authorAntonio Borneo <borneo.antonio@gmail.com>
Sat, 4 Dec 2021 17:38:56 +0000 (18:38 +0100)
committerAntonio Borneo <borneo.antonio@gmail.com>
Sat, 18 Dec 2021 17:26:36 +0000 (17:26 +0000)
On 2008-03-05, before git's age, commit 6d9501467441 adds a
comment about unobserved ACK supposedly sent by GDB at connection.

The ACK is sent since GDB 3.95 (1999-05-04), but a bug introduced
in GDB 6.5 (2006-06-21) and fixed in GDB 7.0 (2009-10-06) makes
GDB sending the query for "supported packets" before sending the
ACK. Due to the bug, the author of the commit failed to see the
ACK.

Change-Id: I574a8013e7d159d1c71087af83b7c2ce92be86bd
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6769
Tested-by: jenkins
src/server/gdb_server.c

index 0d2d1ae2c307eb18ec3321e59c18b31b65d70e75..7c853730f7ac75f055f6b761880fa8f79db2d126 100644 (file)
@@ -1006,16 +1006,19 @@ static int gdb_new_connection(struct connection *connection)
        breakpoint_clear_target(target);
        watchpoint_clear_target(target);
 
-       /* remove the initial ACK from the incoming buffer */
+       /* Since version 3.95 (gdb-19990504), with the exclusion of 6.5~6.8, GDB
+        * sends an ACK at connection with the following comment in its source code:
+        * "Ack any packet which the remote side has already sent."
+        * LLDB does the same since the first gdb-remote implementation.
+        * Remove the initial ACK from the incoming buffer.
+        */
        retval = gdb_get_char(connection, &initial_ack);
        if (retval != ERROR_OK)
                return retval;
 
-       /* FIX!!!??? would we actually ever receive a + here???
-        * Not observed.
-        */
        if (initial_ack != '+')
                gdb_putback_char(connection, initial_ack);
+
        target_call_event_callbacks(target, TARGET_EVENT_GDB_ATTACH);
 
        if (target->rtos) {