Dear Michael Zaidman,
please STOP posting bease64 encoded stuff!
Please send plain text only.
In message 660c0f821002160043g73b85a1bn3b372c9f051a2ce3@mail.gmail.com you wrote:
Working on the POST for mpc834x based board I encountered  the following problem: The ctrlc() routine does not work when used from post_hotkeys_pressed().
Hm... is this not to be expected?
As mentioned before, the POST code is supposed to be NOT interactive. In such an envrionment, handling of ^C makes no sense.
Also, lot of it is running before relocation to RAM, where features like switchable devices or interactive input is not available yet.
The value of ctrlc_disabled variable defined as static in the console.c file is lost after code relocation.
Actually you cannot modify it before relocation.
By adding the ctrlc_disabled to global data gd structure the problem was solved.
It makes no sense to add all sort of random data to gd. gd should be kept strictly minimal.
My questions are:
- I tested this change for ppc branch only. There are a number of boards
belonging to other CPU architectures that also use ctrlc() call from within the post_hotkeys_pressed(). However, I am not sure they use POST at all. I went through the list and found only one such board with CONFIG_POST defined. Does anybody has objection that this change will be expanded to other CPU architectures also?
I object against this change in general. It violates the whole philosophy of POST, which is supposed to be automatic tests.
Best regards,
Wolfgang Denk