nanogui: Thread: running nano-x

Hello,

I'm trying to run nano-x server on my arm platform. I compiled and 
copied bin and lib directories to target fs. After running nano-x I get 
this output:
# ./nano-X
createfont: (height == 0) found builtin font System (0)
createfont: (height == 0) found builtin font System (0)
Select() call in main failed
Select() call in main failed
Select() call in main failed
...

I'm using no keyboard and tuxmouse (ucb1x00 ts) on my target.
I had to modify drivers/Object.rules:
ifeq ($(TUXMOUSE), Y)
CFLAGS += -DTOUCHSCREEN_TUXSCREEN=1
MW_CORE_OBJS += $(MW_DIR_OBJ)/drivers/mou_touchscreen.o
endif

still I get this output while compiling:

Compiling drivers/mou_touchscreen.c ...
/home/xtrm/delo/xxcep/apps/mw-cvs/microwin/src/drivers/mou_touchscreen.c:131: 
warning: initialization from incompatible pointer type
/home/xtrm/delo/xxcep/apps/mw-cvs/microwin/src/drivers/mou_touchscreen.c:50:2: 
#error "You didn't define a device for the generic touchscreen driver!"

I think ts compiler warniings are related to nano-x start failure.

regards,
h
-- 
hinko <dot> kocevar <at> iskramedical <dot> si
Hinko Kocevar, developer
Iskra Medical d.o.o., Stegne 23, 1k LJ, SLO-EU

	"Aì rén"	|	[Analects XII:22]

Hinko Kocevar wrote:
> Hello,
> 
> I'm trying to run nano-x server on my arm platform. I compiled and 
> copied bin and lib directories to target fs. After running nano-x I get 
> this output:
> # ./nano-X
> createfont: (height == 0) found builtin font System (0)
> createfont: (height == 0) found builtin font System (0)
> Select() call in main failed
> Select() call in main failed
> Select() call in main failed
> ...
> 
> I'm using no keyboard and tuxmouse (ucb1x00 ts) on my target.
> I had to modify drivers/Object.rules:
> ifeq ($(TUXMOUSE), Y)
> CFLAGS += -DTOUCHSCREEN_TUXSCREEN=1
> MW_CORE_OBJS += $(MW_DIR_OBJ)/drivers/mou_touchscreen.o
> endif
> 
> still I get this output while compiling:
> 
> Compiling drivers/mou_touchscreen.c ...
> /home/xtrm/delo/xxcep/apps/mw-cvs/microwin/src/drivers/mou_touchscreen.c:131: 
> warning: initialization from incompatible pointer type
> /home/xtrm/delo/xxcep/apps/mw-cvs/microwin/src/drivers/mou_touchscreen.c:50:2: 
> #error "You didn't define a device for the generic touchscreen driver!"
> 
> I think ts compiler warniings are related to nano-x start failure.
> 
> regards,
> h

Ok, I've tried without keyboard and mouse support - the effect is the 
same - Select() call in main failed -

It is OK to compile without keyb and mouse, right? I mean it looks like 
select() fails 'cause of other issues...

-- 
hinko <dot> kocevar <at> iskramedical <dot> si
Hinko Kocevar, developer
Iskra Medical d.o.o., Stegne 23, 1k LJ, SLO-EU

	"Aì rén"	|	[Analects XII:22]

: #error "You didn't define a device for the generic touchscreen driver!"
: 
: I think ts compiler warniings are related to nano-x start failure.

You should set NOMOUSE=Y in config file, and the linked
mouse driver should be drivers/mou_none.c, not the ts driver.

Regards,

Greg

Greg Haerr wrote:
> : #error "You didn't define a device for the generic touchscreen driver!"
> : 
> : I think ts compiler warniings are related to nano-x start failure.
> 
> You should set NOMOUSE=Y in config file, and the linked
> mouse driver should be drivers/mou_none.c, not the ts driver.
> 

I did that. Anyway that doesn't seem to be the case with my problem. I 
tried running the same nano-X on ipaq and it worked flawlessly.

So, the input is not to blame. At startup I see tux on my screen and 
executing cat /dev/fb0 > file and vice-versa also works.

Is this fb funcionality sufficient for nano-x?

I added some printf lines to srvmain.c and found out that errno returned 
after first loop is always 22 - EINVAL; select() man page says that it 
is invalid nr of fds passed to select(), but without keyb and mouse 
setsize = 3.

regards,
h
-- 
hinko <dot> kocevar <at> iskramedical <dot> si
Hinko Kocevar, developer
Iskra Medical d.o.o., Stegne 23, 1k LJ, SLO-EU

	"Aì rén"	|	[Analects XII:22]

: So, the input is not to blame. At startup I see tux on my screen and 
: executing cat /dev/fb0 > file and vice-versa also works.
: 
: Is this fb funcionality sufficient for nano-x?

Yes, should be


: 
: I added some printf lines to srvmain.c and found out that errno returned 
: after first loop is always 22 - EINVAL; select() man page says that it 
: is invalid nr of fds passed to select(), but without keyb and mouse 
: setsize = 3.

Try printf()ing the values of the 3 fds that are passed, it may be
that the mousefd or keyfd are negative...

Regards,

Greg

Greg Haerr wrote:

> Try printf()ing the values of the 3 fds that are passed, it may be
> that the mousefd or keyfd are negative...

mouse_fd is -2 if print()ed in GsInitialise() - result of NUL_Open(). 
Same goes for keyb_fd which is -2 also for the same reason.

In GsSelect() those fds don't get in fd_set, since they are not >= 0. 
un_sock takes up fb nr. 3. These are print()outs of fds in fd_Set just 
before call to select() is made:

maxfd = setsize+1;
for(i=0; i<=maxfd; i++)
           printf("%d\t%d\n", i, FD_ISSET(i,&rfds));

FD     val
0	0
1       0
2       0
3       1
4       0



-- 
hinko <dot> kocevar <at> iskramedical <dot> si
Hinko Kocevar, developer
Iskra Medical d.o.o., Stegne 23, 1k LJ, SLO-EU

	"Aì rén"	|	[Analects XII:22]

> : I added some printf lines to srvmain.c and found out that errno returned
> : after first loop is always 22 - EINVAL; select() man page says that it
> : is invalid nr of fds passed to select(), but without keyb and mouse
> : setsize = 3.

Perhaps the select() is failing because your kernel
isn't configured for sockets or networking?

If you configure mwin with the basic tty driver, does
this still happen?

I'm thinking that since setsize=3, and un_sock is 3,
then there's an issue with the kernel complaining
about selecting on that (the client accept) socket descriptor.

Regards,

Greg

Greg Haerr wrote:
>>: I added some printf lines to srvmain.c and found out that errno returned
>>: after first loop is always 22 - EINVAL; select() man page says that it
>>: is invalid nr of fds passed to select(), but without keyb and mouse
>>: setsize = 3.
> 
> 
> Perhaps the select() is failing because your kernel
> isn't configured for sockets or networking?
> 
my .config in 2.4.19 says
...
CONFIG_UNIX=y
CONFIG_INET=y
...

> If you configure mwin with the basic tty driver, does
> this still happen?
> 

Will try...

> I'm thinking that since setsize=3, and un_sock is 3,
> then there's an issue with the kernel complaining
> about selecting on that (the client accept) socket descriptor.
> 

setsize is set only here
	if (un_sock > setsize) setsize = un_sock;

and additionaly if client connects, here
	if(curclient->id > setsize) setsize = curclient->id;

and setsize+1 is later passed to select()...


I ran strace on nano-X and this line does not seem right:

select(4, [3], NULL, NULL, {3813670912, 3785383936}) = -1 EINVAL 
(Invalid argument)

-> exceptfds and timeout are bogus !?

I belive it should read
select(4, [3], NULL, NULL, NULL)

regards,
h

-- 
hinko <dot> kocevar <at> iskramedical <dot> si
Hinko Kocevar, developer
Iskra Medical d.o.o., Stegne 23, 1k LJ, SLO-EU

	"Aì rén"	|	[Analects XII:22]

> I ran strace on nano-X and this line does not seem right:
> 
> select(4, [3], NULL, NULL, {3813670912, 3785383936}) = -1 EINVAL 
> (Invalid argument)
> 
> -> exceptfds and timeout are bogus !?
> 
> I belive it should read
> select(4, [3], NULL, NULL, NULL)

No, the last parameter is the timeout struct, which is
required for timing out of the select().  As I mentioned
in an earlier post, I'd bet the structure/size is different
between kernel and/or compiler versions.  Also, earlier
versions of ARM didn't default to 32-bit structure size
padding, like Intel (this caused my first Doom port to
iPAQ to fail, since Doom won't run without 32-bit
structure padding ala x86)

Regards,

Greg

I tried to run nano-X but I get an error:

Cannot bind to named socket
Couldn't connect to Nano-X Server!


What should I do ?