[BUG] sandbox: './u-boot -l ' fails

Tue Sep 29 01:59:07 CEST 2020

On 9/29/20 12:19 AM, Simon Glass wrote:
> Hi Heinrich,
>
> On Mon, 28 Sep 2020 at 15:23, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
>>
>> On 9/28/20 3:42 PM, Simon Glass wrote:
>>> Hi Heinrich,
>>>
>>> On Mon, 28 Sep 2020 at 07:30, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
>>>>
>>>> On 28.09.20 15:22, Simon Glass wrote:
>>>>> Hi Heinrich,
>>>>>
>>>>> On Mon, 28 Sep 2020 at 05:31, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
>>>>>>
>>>>>> On 28.09.20 06:46, Heinrich Schuchardt wrote:
>>>>>>> Am 28. September 2020 06:24:38 MESZ schrieb Simon Glass <sjg at chromium.org>:
>>>>>>>> Hi Heinrich,
>>>>>>>>
>>>>>>>> On Sat, 19 Sep 2020 at 13:48, Heinrich Schuchardt <xypron.glpk at gmx.de>
>>>>>>>> wrote:
>>>>>>>>>
>>>>>>>>> Hello Simon,
>>>>>>>>>
>>>>>>>>> when I try to run ./u-boot -l the sandbox stalls. Shouldn't it run
>>>>>>>> out
>>>>>>>>> of the box?
>>>>>>>>>
>>>>>>>>> $ ./u-boot -l -d arch/sandbox/dts/sandbox.dtb
>>>>>>>>
>>>>>>>> For the record you should be able to use -D to get the same effect as
>>>>>>>> your -d above.
>>>>>>>>
>>>>>>>>>
>>>>>>>>> U-Boot 2020.10-rc4-00018-g21a10244f9-dirty (Sep 19 2020 - 19:55:39
>>>>>>>> +0200)
>>>>>>>>>
>>>>>>>>> Model: sandbox
>>>>>>>>> DRAM:  128 MiB
>>>>>>>>>
>>>>>>>>> Warning: host_lo MAC addresses don't match:
>>>>>>>>> Address in ROM is               26:4b:ca:6c:98:f4
>>>>>>>>> Address in environment is       00:00:11:22:33:44
>>>>>>>>>
>>>>>>>>> Warning: host_virbr0 MAC addresses don't match:
>>>>>>>>> Address in ROM is               ee:3e:c9:ce:1f:9c
>>>>>>>>> Address in environment is       00:00:11:22:33:45
>>>>>>>>>
>>>>>>>>> Warning: host_docker0 MAC addresses don't match:
>>>>>>>>> Address in ROM is               c2:85:07:7b:9a:18
>>>>>>>>> Address in environment is       00:00:11:22:33:46
>>>>>>>>> WDT:   Not found!
>>>>>>>>> MMC:
>>>>>>>>>
>>>>>>>>> No output after this point.
>>>>>>>>>
>>>>>>>>> The problem also exists with U-Boot v2020.07, v2019.10, v2018.11.
>>>>>>>>>
>>>>>>>>> CONFIG_SANDBOX_SDL=y
>>>>>>>>>
>>>>>>>>> SDL_InitSubSystem() never returns. It is looping somewhere in
>>>>>>>> U-Boot's
>>>>>>>>> __serial_getc(). I wonder how it gets there without returning from
>>>>>>>> the
>>>>>>>>> function.
>>>>>>>>>
>>>>>>>>> I compiled SDL2.cpp from
>>>>>>>>> https://gist.github.com/miguelmartin75/6946310#file-sdl2-cpp-L18
>>>>>>>>> with
>>>>>>>>>
>>>>>>>>> g++ SDL2.cpp -D_REENTRANT -I/usr/include/SDL2 -lSDL2 -lGL -o test
>>>>>>>>>
>>>>>>>>> and it runs fine showing an X11 windows with red background.
>>>>>>>>>
>>>>>>>>> So there seems to be no general problem with the SDL2 library.
>>>>>>>>
>>>>>>>> I hit this myself on another computer and it turned out to be that SDL
>>>>>>>> defined getc(), as does U-Boot, and things get confused. At least I
>>>>>>>> think it is getc.
>>>>>>>>
>>>>>>>> I hacked around with changing the name of getc (I think it was getc)
>>>>>>>> in U-Boot and the problem went away.
>>>>>>>
>>>>>>> Should we include a patched SDL2 with U-Boot for static linking?
>>>>>>>
>>>>>>
>>>>>> I cannot find a symbol getc() nor a reference to it in the SDL
>>>>>> repository. getc() is defined in stdio.h.
>>>>>>
>>>>>> Our getc() takes no argument, while the stdio one wants a FILE *.
>>>>>>
>>>>>> But changing the U-Boot definition to "int getc(void *)" does not solve
>>>>>> the issue.
>>>>>
>>>>> I just tried it on the machine where it doesn't work:
>>>>>
>>>>> $ gdb --args /tmp/b/sandbox/u-boot -l -D
>>>>> GNU gdb (Debian 9.2-1) 9.2
>>>>> Copyright (C) 2020 Free Software Foundation, Inc.
>>>>> License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
>>>>> This is free software: you are free to change and redistribute it.
>>>>> There is NO WARRANTY, to the extent permitted by law.
>>>>> Type "show copying" and "show warranty" for details.
>>>>> This GDB was configured as "x86_64-linux-gnu".
>>>>> Type "show configuration" for configuration details.
>>>>> For bug reporting instructions, please see:
>>>>> <http://www.gnu.org/software/gdb/bugs/>.
>>>>> Find the GDB manual and other documentation resources online at:
>>>>>     <http://www.gnu.org/software/gdb/documentation/>.
>>>>>
>>>>> For help, type "help".
>>>>> Type "apropos word" to search for commands related to "word"...
>>>>> Reading symbols from /tmp/b/sandbox/u-boot...
>>>>> (gdb) br getc
>>>>> Breakpoint 1 at 0x5b6d6: getc. (2 locations)
>>>>> (gdb) r
>>>>> Starting program: /tmp/b/sandbox/u-boot -l -D
>>>>> [Thread debugging using libthread_db enabled]
>>>>> Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
>>>>>
>>>>>
>>>>> U-Boot 2020.10-rc5-00196-g0ac83d080a0 (Sep 28 2020 - 07:11:52 -0600)
>>>>>
>>>>> Model: sandbox
>>>>> DRAM:  128 MiB
>>>>>
>>>>> Warning: host_lo MAC addresses don't match:
>>>>> Address in ROM is 1a:34:9b:48:aa:53
>>>>> Address in environment is 00:00:11:22:33:44
>>>>> WDT:   Not found!
>>>>> MMC:
>>>>>
>>>>> Breakpoint 1, getc () at
>>>>> /home/sjg/cosarm/src/third_party/u-boot/files/common/console.c:414
>>>>> 414 if (!gd->have_console)
>>>>> (gdb) up
>>>>> #1  0x00007ffff7914d48 in ?? () from /lib/x86_64-linux-gnu/libX11.so.6
>>>>> (gdb)
>>>>>
>>>>>
>>>>> So it seems that it is libX11 causing the problem.
>>>>>
>>>>> I don't think the best fix is to remove getc() from that library,
>>>>> although I do wonder if it is a bug. Renaming the symbol in U-Boot
>>>>> with #define (only on sandbox) might be one option.
>>>>>
>>>>> Regards,
>>>>> Simon
>>>>>
>>>>
>>>> #define getc _uboot_getc
>>>>
>>>> is what I tried but it runs you into the trouble that many other symbols
>>>> contain the getc substring.
>>>>
>>>> As we always try to use the same definition in U-Boot as in glibc we
>>>> should really replace getc() by another symbol, e.g. chget() so that we
>>>> avoid confusion.
>>>
>>> Well the first question is whether this is a bug with x11.
>>>
>>> I'm not sure that renaming is a good idea. getc() is a pretty standard name.
>>>
>>> Try
>>>
>>> #define getc(x) _sandbox_get(x)
>>>
>>> which might get you closer.
>>>
>>> Regards,
>>> Simon
>>>
>> https://gitlab.denx.de/u-boot/custodians/u-boot-efi/-/commit/7b302bedb21f94b9eb619cb38ef985f7bbb9be1f
>>
>> seems to fix the crash. I get an output window.
>>
>> But there are still problems:
>>
>> * The shift key does not work on the keyboard.
>
> What is shift supposed to do? Do you mean you cannot get capital
> letters or symbols?
>
>> * The output is black and white only.
>
> That's odd. I haven't tried it recently but earlier in the year I was
> able to write a colour BMP.
>
>>
>> What I would like to have is a 32bit color framebuffer on which I can
>> draw with the EFI graphical output protocol.
>
> Well that should work. I adds Nuklear support a while back and
> everything seemed good. Perhaps there has been a regression? But there
> is a test for colour bitmaps. Perhaps it is just the SDL display side
> which is not working?
>
> I just tried this and got a colour image:
>
> => host load hostfs - 0 tools/logos/denx.bmp
> => bmp display 0
>
> Regards,
> Simon
>

Hello Simon,

the background of characters is not set correctly. I thought it was
black and white.

To see what I mean:

CONFIG_EFI_SELFTEST=y

./u-boot -D -l

=> setenv efi_selftest block image transfer
=> bootefi selftest

You will need my two patches.

Things that are still wrong:

* We start with black on white instead white on black.
* Background for characters.
* The font in not mono-spaced.
* When typing a backslash you get always two of them.

Best regards

Heinrich