Dobrica Pavlinušić's random unstructured stuff
HTC Dream G1 JTAG: Revision 5

Debugging

http://forum.xda-developers.com/showpost.php?p=6498820&postcount=621

That said before doing anything else take out your multi meter (and if you don't have one you are missing a tool for this type of work) and check the following:

blue-light mode

1) Put phone into blue light mode if serial is attached and power isn't: you will see bootmode 1

dpavlin@x200:/virtual/android$ ./neocon /dev/ttyUSB0 
[Closed]
[Open /dev/ttyUSB0]

boot reason: PM_KPD_PWR_KEY_ON_RT_ST

(PowerOn Status,Boot Reason)=(1,1)
NAND_FLASH_READ_ID : SAMSUNG_256MB_FLASH_128MB_SDRAM

ARM9_BOOT_MODE1
Invalid command : ?

v-ref = 2.6v

2) verify your v-ref is actually 2.6v (usually within 0.05v) when compared to the ground (any of the shielding) of the main board. You have one of the right points so there is an issue with the connection if its not.

oprnocd, trst-n = 2.6v

3) now start open ocd: and check that trst-n is now also 2.6v when compared to ground. (if not your adapter is not working with the 2.6v.. very possible)

nTRST must be at high level (~2.6V) after openocd is launched.
With nTRST sticking low the MSM7201A debug unit is in reset state.

Maybe you'll have to tweak the cfg file. There are several options for the behaviour of nTRST.
If you don't manage to find a working cfg, you may cut the nTRST connection to your adaptor and pull the Dream's nTRST signal to Vref=2.6V permanently.
Normally this should also work... the debug unit will then leave reset state immediately after power up.

Maybe you'll have to tweak the cfg file. There are several options for the behaviour of nTRST.
If you don't manage to find a working cfg, you may cut the nTRST connection to your adaptor and pull the Dream's nTRST signal to Vref=2.6V permanently.
Normally this should also work... the debug unit will then leave reset state immediately after power up.

soldering

4) with that done and you are still having problems as we said before check the soldering work.. Here there are two possibilities:

    A) bridges - the wire is in contact with something in addition to the testpoint
    B) bad joints - while the wire may act attached it is not. (Others must be able to explain this better than myself..)

    The tiniest bit of flux goes a long way here..

    One quick thing you can test is that none of the 5 test points are connected to ground (disconnect rtck for now its one less variable) and that none are shorted to eachother.

other

Other things to check:

  • How long are the wires.. My setup runs much faster and the wires are not exactly short .. but the more wire the more chance for noise..
  • is the speed acceptable with the parport (if this is a real parport I'll hope openocd has sane defaults but it is something to keep in mind)
  • phone is in blue light mode (the phone can disable the jtag port.. this happens when amss is booted either via the GO2AMSS command or when the linux kernel is started. I also have no issue connecting to JTAG while the battery is charging.
  • Ensure the openocd application is not running when you boot the phone.