m-labs / bugs Goto Github PK

To reproduce the problem:

[/] # ls
dev flash ramdisk
[/] # cd flash
[/flash] # ls
lost+found patches sysconfig.bin wallpapers
[/flash] # cd wallpapers
[/flash/wallpapers] # ls
Comet.png Curlybracket-light-04.png Wall_paper_1.png
[/flash/wallpapers] # rm Wall_paper_1.png
[/flash/wallpapers] # ls
Comet.png Curlybracket-light-04.png
[/flash/wallpapers] # rm Curlybracket-light-04.png
[/flash/wallpapers] # ls
Comet.png
[/flash/wallpapers] # mkdir a
[/flash/wallpapers] # mkdir a/b
mkdir 'a/b' failed:No such file or directory
[/flash/wallpapers] # cd a
[/flash/wallpapers/a] # mkdir b
[/flash/wallpapers/a] # cd b
[/flash/wallpapers/a/b] # mkdir c
[/flash/wallpapers/a/b] # cd c
[/flash/wallpapers/a/b/c] # cp /flash/wallpapers/Comet.png ./a.png
./a.png: No such file or directory
[/flash/wallpapers/a/b/c] # cp /flash/wallpapers/Comet.png a.png
[/flash/wallpapers/a/b/c] # ls
a.png
[/flash/wallpapers/a/b/c] # cd /flash
[/flash] # ls
lost+found patches sysconfig.bin wallpapers
[/flash] # ls wallpapers/a/b/c/
a.png
[/flash] #

Characters like ô, è, ê, é are not displayed correctly in the window title font.

The WM9707 chip we are using has a "mic boost" option that adds +20dB. Add support in the RTEMS driver and a button in the audio settings to use it.

Divide key should print "/" instead of the division symbol. The division symbol is useless, FNP and URLs use "/".

Moving files between flash and ramdisk does not work. It copies empty files instead.

MDIO would enable checking the link status and avoid waiting for DHCP when the Ethernet cable is disconnected.

could you add GDB to the toolchain built by the scripts?

For GDB < 7.3 you'll need this patch for the serial cable to work correctly:
http://sourceware.org/ml/gdb-patches/2011-04/msg00066.html

The CRT of the debug stub should push IE, IP and IM on the stack for GDB to retrieve later.

Need to add some ability to set a range for the changing of patches so its not always so easily predictable the change every 30 seconds. Also great to be able to set some long times. I think by default should have 1.5 - 3 minute changes to generically match songs. Need GUI elements for this...and should have sensibles default.

supposed fix:

--- a/software/gdbstub/gdbstub.c
+++ b/software/gdbstub/gdbstub.c
@@ -505,7 +505,7 @@ static void cmd_mem_read(void)

 /* try to read %x,%x */
 if (hex2int(&ptr, &addr) > 0 && *ptr++ == ',' && hex2int(&ptr, &length) > 0

•        && length < (sizeof(remcom_out_buffer) / 2)) {
  

•        && length <= sizeof(remcom_out_buffer) / 2) {
   if (mem2hex((char *)addr, remcom_out_buffer, length) == NULL) {
       strcpy(remcom_out_buffer, "E14");
   }
  

When you issue a lot of consecutive commands to MTK (e.g. during application startup), it ends up not redrawing anything except when new events come in.

A workaround is to issue a screen.refresh command when the problem appears.

Pressing and holding the backspace key does not repeatedly delete characters. Autorepeat works for other keys including delete.

1. when boot with camera running, it is always detected, but when I disconnect/reconnect during runtime, chances are good that I will 'loose' the camera, until I reboot
2. after system boot. then connect the camera, sometime system can not detect the camera

Build: 1.1-final

Description: When use a image [1] with color text on white background in a patch, the color on text look brushed off when the patch is rendered. Use the same image as wallpaper will have the same result.

Reproduce steps:

1. Transfer file [1] to M1 /ssd/patchpool/ folder
2. Create a new patch include the following lines:
   imagefile1=color_white_bg.jpg
   decay=0
   image1_a=1
3. Save and Run the patch.
   => Notice the text look like brushed [2]

[1] http://en.qi-hardware.com/wiki/File:Color_white_bg.jpg
[2] http://en.qi-hardware.com/wiki/File:Color_white_bg_screenshot.png

by using the new minimac2, my m1 ethernet
no long working under my 8-ports 10BAST-T hub(TP-LINK), DIR-300(D-LINK home router) DIR-523M(D-LINK home router) when I connect m1 to those device. the two leds of ethernet are off.

but it's working when I direct connect the m1 to my laptop(100M/1000M
Ethernet controller: nVidia Corporation MCP79 Ethernet (rev b1)) it working in 100-BASE-T 24 port switcher two.

The softusb-input firmware currently only supports keyboards and mice with only one interface and one endpoint. This causes (rare) keyboard + mouse combo devices not to work, e.g.

Bus 003 Device 008: ID 04d9:2809 Holtek Semiconductor, Inc. 
Device Descriptor:
  bLength                18
  bDescriptorType         1
  bcdUSB               1.10
  bDeviceClass            0 (Defined at Interface level)
  bDeviceSubClass         0 
  bDeviceProtocol         0 
  bMaxPacketSize0         8
  idVendor           0x04d9 Holtek Semiconductor, Inc.
  idProduct          0x2809 
  bcdDevice            1.10
  iManufacturer           1  
  iProduct                2 USB Keyboard
  iSerial                 0 
  bNumConfigurations      1
  Configuration Descriptor:
    bLength                 9
    bDescriptorType         2
    wTotalLength           59
    bNumInterfaces          2
    bConfigurationValue     1
    iConfiguration          0 
    bmAttributes         0xa0
      (Bus Powered)
      Remote Wakeup
    MaxPower              100mA
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        0
      bAlternateSetting       0
      bNumEndpoints           1
      bInterfaceClass         3 Human Interface Device
      bInterfaceSubClass      1 Boot Interface Subclass
      bInterfaceProtocol      1 Keyboard
      iInterface              0 
        HID Device Descriptor:
          bLength                 9
          bDescriptorType        33
          bcdHID               1.10
          bCountryCode            0 Not supported
          bNumDescriptors         1
          bDescriptorType        34 Report
          wDescriptorLength      62
         Report Descriptors: 
           ** UNAVAILABLE **
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x81  EP 1 IN
        bmAttributes            3
          Transfer Type            Interrupt
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0008  1x 8 bytes
        bInterval              10
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        1
      bAlternateSetting       0
      bNumEndpoints           1
      bInterfaceClass         3 Human Interface Device
      bInterfaceSubClass      1 Boot Interface Subclass
      bInterfaceProtocol      2 Mouse
      iInterface              0 
        HID Device Descriptor:
          bLength                 9
          bDescriptorType        33
          bcdHID               1.10
          bCountryCode            0 Not supported
          bNumDescriptors         1
          bDescriptorType        34 Report
          wDescriptorLength     160
         Report Descriptors: 
           ** UNAVAILABLE **
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x82  EP 2 IN
        bmAttributes            3
          Transfer Type            Interrupt
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0008  1x 8 bytes
        bInterval              10
Device Status:     0x0000
  (Bus Powered)

RTEMS can send its syslog messages over Ethernet, apparently with a protocol close to (or compliant with) RFC 5424. This is an undocumented feature of RTEMS.

Simply setting the log_host parameter to an IP address string in the rtems_bsdnet_config structure does not work. The UDP packet is apparently sent in the vsyslog function from cpukit/libnetworking/lib/syslog.c.

This would enable a printf-like feature over Ethernet, without the need to open the box to connect a serial cable.

• Need to write INSEL register (see ADV7181 datasheet p.14) to select between the three
• Add ioctl in RTEMS driver to do that
• Add a 3-position radiobutton in the video input settings in Flickernoise
• Read/write to performance config (like contrast, brightness and hue)

Instead of hardcoding the CRC and length in the source code (which will be messy), you can write it at the end of the boot.bin e.g. using echo and dd (like you build the rescue BIOS images with the MAC addresses). You can have a look at how the BIOS retrieves its own CRC and do the same.

How to reproduce:

• Power on M1
• In Flickernoise, click Shutown->Power off
• Immediately press the middle pushbutton to power on again

Symptoms:

• Complete loss of Ethernet functionality
• LEDs on the Ethernet connector are always on, even in the absence of a cable

Probably, the signals sent from the FPGA when reloading bitstreams confuse the PHY chip and make it enter a "broken" state. For some reason, the reset pulse which is supposedly sent immediately after the loading of the main bitstream fails to put the PHY back into a normal state. The problem goes away after a second soft system (and PHY) reset.

Make it clear that those includes come from MTK, also to avoid namespace conflicts. Add a "mtk" prefix to their file names.

The compiler temporarily preallocates PFPU registers for all potential variables (including unused ones) before starting to compile a patch. It then allocates additional registers for constants and user variables during the patch's compilation. Lately, I have added 16 more variables (4 more DMX in, 4 more DMX out and 8 MIDI inputs) which means there are 16 less registers available for constants and user variables. This caused a few patches, which use a great amount of constants, to fail at compilation.

The proper way to fix it would be to have a smarter allocation algorithm that would scan the patch before compilation and only preallocate registers for variables that are actually used.

Symptom: libcurl connect timeout does not work
Using CURLOPT_CONNECTTIMEOUT and CURLOPT_TIMEOUT has no effect. Transfers that should be interrupted by timeouts are not.
Those features are needed in Flickernoise so that the render mode can be cleanly exited when the RSS wall is enabled with a server that does not respond.
See commit m-labs/flickernoise@1c8d4a5

Autostart seems to crash the system when the GUI resolution is 640x480. We should:
a) fix this bug
b) disable loading of the sysconfig.bin from the flash when booted in rescue mode, so that other potential problems like this do not end up with an unrecoverable state for the user.

Sometimes the MDIO read function in libnet does not get the right value, e.g. 0x0045 instead of 0x0022. This probably has to do with sampling the data at the wrong time in the bit banging functions:
0x0045 = (0x0022 << 1) | 0x0001

The '?' key does nothing instead of writing a character, at least on the German keyboard.

after boot to rescue mode(hold left button then press middle button will boot to rescue mode). I have to un-plug and plug power cable again for normal boot

e.g. a, q, and 0 to 9 (and the symbols on those keys). This is the last thing to fix before 0.4.

The screen gets tinted in blue when the OSD layer is displayed in QEMU. Probably a bug with the way chroma key is implemented in the TMU model.

Normal behaviour:
[/] # cd ramdisk
[/ramdisk] # mkdir blah
[/ramdisk] # mkdir /ramdisk/blah
mkdir '/ramdisk/blah' failed:File exists
[/ramdisk] #
[/ramdisk] # mkdir /ramdisk/blah/
mkdir '/ramdisk/blah/' failed:File exists

Incorrect behaviour:
[/] # cd ssd
[/ssd] # mkdir blah
[/ssd] # mkdir /ssd/blah
mkdir '/ssd/blah' failed:File exists
[/ssd] # mkdir /ssd/blah/
[/ssd] # ls blah
obj772
[/ssd] #

like: finder, filezilla, nautilus etc. it give you strange error. more info: http://en.qi-hardware.com/mmlogs/latest.log.html#t04:47

Remove unneeded key codes, add relevant layout-independent key codes, and do all the keymap processing in input.c.

set RTEMS FTPd "root directory" to /ssd, and also hide non-ssd directories in the GUI

Sebastien said the RTEMS has encoders so we can add audio playback, which would be great for having a standalone music player in the milkymist for demos.

Which kinda defeats its purpose.

RTEMS features a GDB server that one could enable and use to debug over Ethernet. Let's get that to work in order to enable debugging some cases without opening the box.