diff --git a/ChangeLog b/ChangeLog
index f6e64259c2dae1dcc62012446ea8992afd3f2b45..61b4d5e3ab19e798fc1ed6c583fb636d29b8d733 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -567,4 +567,8 @@
* Add poll() and select() APIs (in the initial check-in, these work only with character devices)
* Add poll() methods to /dev/null, /dev/zero, pipes, fifos, and serial drivers.
* Add examples/poll for testing poll() and select()
+ * Fix hostile behavior of getc, fgetc, getchar, etc.: the serial driver was waiting for a
+ full buffer of read data before return. This means that getc would stall when it needed
+ to refill the input buffer. The old behavior (read full blocks) might be useful in other
+ contexts, so it is still available within the driver as a configuration option.
diff --git a/Documentation/NuttX.html b/Documentation/NuttX.html
index cf3e9603d111460b7fb46a46639147619de885ba..24c3a23a354ed93ca9f52bfd299ca3151c43484d 100644
--- a/Documentation/NuttX.html
+++ b/Documentation/NuttX.html
@@ -1203,6 +1203,10 @@ nuttx-0.3.19 2008-xx-xx Gregory Nutt <spudmonkey@racsa.co.cr>
* Add poll() and select() APIs (in the initial check-in, these work only with character devices)
* Add poll() methods to /dev/null, /dev/zero, pipes, fifos, and serial drivers.
* Add examples/poll for testing poll() and select()
+ * Fix hostile behavior of getc, fgetc, getchar, etc.: the serial driver was waiting for a
+ full buffer of read data before return. This means that getc would stall when it needed
+ to refill the input buffer. The old behavior (read full blocks) might be useful in other
+ contexts, so it is still available within the driver as a configuration option.
pascal-0.1.3 2008-xx-xx Gregory Nutt <spudmonkey@racsa.co.cr>
diff --git a/TODO b/TODO
index 3475174beb9ffc43835c13991c53c68299cc2863..2f85472e27b6ffeb604ae1ca99a61d917457810a 100644
--- a/TODO
+++ b/TODO
@@ -9,13 +9,14 @@ NuttX TODO List (Last updated November 17, 2008)
(1) C++ Support
(12) Network (net/, netutils/)
(1) USB (drivers/usbdev)
- (3) Libraries (lib/)
+ (4) Libraries (lib/)
(6) File system/Generic drivers (fs/, drivers/)
(1) Pascal add-on (pcode/)
(2) Documentation (Documentation/)
(3) Build system
(2) NuttShell (NSH) (examples/nsh)
(1) Other Applications & Tests (examples/)
+ (2) Linux/Cywgin simulation (arch/sim)
(2) ARM (arch/arm/)
(1) ARM/C5471 (arch/arm/src/c5471/)
(1) ARM/DM320 (arch/arm/src/dm320/)
@@ -230,13 +231,19 @@ o Libraries (lib/)
Status: Open
Priority: Medium
- Description: Buffered C, character-oriented I/O is not very useful when
- dealing with a console. The behavior is to read a full buffer
- of data then return the individual characters from the buffer.
- But that is very hostile when working with a human interface.
+ Description: fgets implementation does not use C-buffered I/O, but rather
+ talks to serial driver directly via read(). It includes VT-100
+ specific editting commands. This gets should be renamed readlin()
+ and a more generic fgets() should be implemented.
Status: Open
- Priority: Medium. Workarounds: Set CONFIG_STDIO_BUFFER_SIZE=0 in the
- configuration file.
+ Priority: Low (unless you are using mixed C-buffered I/O with fgets and
+ fgetc, for example).
+
+ Description: Need some minimal termios support... at a minimum, enough to
+ switch between raw and "normal" modes to support behavior like
+ that needed for readline().
+ Status: Open
+ Priority: Low
o File system / Generic drivers (fs/, drivers/)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -318,16 +325,21 @@ o Build system
Priority: Low
Description: Dependencies do not work correctly under configs/<board>/src
- (same as arch/<arch>/src/board). Seems to be worse using SDCC.
+ (same as arch/<arch>/src/board).
Status: Open
Priority: Medium (maybe higher for z80 target)
Description: It would be nice to de-couple some of the behaviors enabled by
CONFIG_DEBUG. For example, CONFIG_DEBUG turns on debug output,
but also disables optimization make the binary much bigger.
- Status: Low. There workaround is easy.. just edit the configuration's
+ Status: Open
+ Priority: Low. There workaround is easy.. just edit the configuration's
Make.def file to control the behavior that you want.
- Priority: Very low.
+
+ Description: Need a NuttX configuration tool. The number of configuration
+ settings has become quite large and difficult to manage manually.
+ Status: Open
+ Priority: Medium-low
o NuttShell (NSH) (examples/nsh)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -351,6 +363,24 @@ o Other Applications & Tests (examples/)
incorrectly on the Cywgin-based simulation platform (but works
fine on the Linux-based simulation platform).
+o Linux/Cywgin simulation (arch/sim)
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+ Description: Need to implement timing. Use system timing resources to fake
+ a NuttX real time clock. At present, the timer is driven in
+ non-realtime (like 1000x realtime) and causes some behavior
+ differences from real platforms.
+ Status: Open
+ Priority: Low (because the simulator is only a test/development platform)
+
+ Description: The simulated serial driver also has some odd behavior. It
+ will stall for a long time on reads when the C stdio buffers are
+ being refilled. This only effects the behavior of things like
+ fgetc(). Workaround: Set CONFIG_STDIO_BUFFER_SIZE=0, suppressing
+ all C buffered I/O.
+ Status: Open
+ Priority: Low (because the simulator is only a test/development platform)
+
o ARM (arch/arm/)
^^^^^^^^^^^^^^^
diff --git a/drivers/serial.c b/drivers/serial.c
index 2939bf61f00b2e8dc86aa9b26ffc0506926b65d3..533675886dd12493e29dd8501f50eb0590988053 100644
--- a/drivers/serial.c
+++ b/drivers/serial.c
@@ -318,8 +318,9 @@ static ssize_t uart_read(FAR struct file *filep, FAR char *buffer, size_t buflen
}
}
+#ifdef CONFIG_DEV_SERIAL_FULLBLOCKS
/* No... then we would have to wait to get receive more data.
- * If the user has specified the O_NONBLOCK option, then do not
+ * If the user has specified the O_NONBLOCK option, then just
* return what we have.
*/
@@ -333,15 +334,36 @@ static ssize_t uart_read(FAR struct file *filep, FAR char *buffer, size_t buflen
{
recvd = -EAGAIN;
}
+ break;
+ }
+#else
+ /* No... the circular buffer is empty. Have we returned anything
+ * to the caller?
+ */
- /* Break out of the loop and return the number of bytes
+ else if (recvd > 0)
+ {
+ /* Yes.. break out of the loop and return the number of bytes
* received up to the wait condition.
*/
break;
- }
+ }
- /* Otherwise we are going to wait */
+ /* No... then we would have to wait to get receive some data.
+ * If the user has specified the O_NONBLOCK option, then do not
+ * wait.
+ */
+
+ else if (filep->f_oflags & O_NONBLOCK)
+ {
+ /* Break out of the loop returning -EAGAIN */
+
+ recvd = -EAGAIN;
+ break;
+ }
+#endif
+ /* Otherwise we are going to have to wait for data to arrive */
else
{
diff --git a/examples/README.txt b/examples/README.txt
index 2d8ea8a2d18e5e6cb27d7ac323dc2f2ba792c4f9..2d16b210fc62eb3717bcfffd0eca119e68c46df0 100644
--- a/examples/README.txt
+++ b/examples/README.txt
@@ -124,8 +124,7 @@ examples/serloop
* CONFIG_EXAMPLES_SERLOOP_BUFIO
Use C buffered I/O (getchar/putchar) vs. raw console I/O
- (read/read). The behavior of the NuttX getchar() call is
- very hostile unless you also set CONFIG_STDIO_BUFFER_SIZE=0.
+ (read/read).
examples/udp
^^^^^^^^^^^^