RE: [ECOS] Nano using PDCurses for eCos

RE: [ECOS] Nano using PDCurses for eCos

[Steven Clugston]

[-- Attachment #1: Type: text/plain, Size: 3352 bytes --]

> If you just want to edit config files, it's better to implement own
> editor, call it `femto' then :-)
> http://en.wikipedia.org/wiki/SI_prefix
> It seems for me that nano is enough bloat stuff for eCos targets.

Nano does far more than I realised and I agree that most of it is bloat for what I need like calls to external programs for spell checking and wide character and NLS support etc, but it's a choice between writing an editor from scratch or hacking nano to work by disabling the bloat using its existing config mechanism. I think the latter should take less time/effort. It's a shame there's not a #define NANO_FEMTO mode to strip off even more than NANY_TINY does.

> Yesterday evening I did grab the latest stable nano sources (2.0.9)
> from here http://www.nano-editor.org, and tried to  > build it. After
> configure and some tweaks I got The GNU nano editor for the eCos
> i386linux target just in 110K using those PDCurses package and dummy
> driver.

> I should not have told about until a try. Steven, I have to finalize
> and test the build (at the least using eCos
> FS_RAM), I have to add the real terminal driver to PDCurses. I do hope this will occur the next weekend and then I will
> share a full story somewere on the web (some guys are nervious if they cannot read here only a digest about eCos kernel
> gears and cpu parts :-) There are a few hints the below

I know this is bordering on another general programming discussion, but its because of the issues which ecos throws up that I'm discussing it and it just might interest some other ecos user out there one day.

> I configured the nano as: --enable-tiny --disable-nls --disable-utf8;
> I have a bit tweaked config.h to point on that
> which things are absent in eCos, made a peace with Linux/GNU (pwd, mkstemp, signals, getopt) and I got the executable.
> There is alone rule in my eCos Makefile

>    ACTUAL_CFLAGS := $(subst , -DHAVE_CONFIG_H=1,$(ACTUAL_CFLAGS))

I've actually got nano 2.0.7 to compile to a binary with modifying any of the actual nano source. It seems to run on my PowerPC board now. I can't tell if it works yet as it doesn't accept any input, but the main screen fires up. For some reason my raw binary is around 413Kb which seems quite big, but I probably could have disabled more stuff in CDL like unneeded hardware packages.

Here's what I did:

I just ran configure on the source in Linux to get the basic config.h file and then modified that to disable most headers and things that are not present in ecos and enabled NANO_TINY as well.

To get the termios stuff and unix signals to compile I had to use the POSIX compatibility layer package and enable POSIX signals and disable the ISO C signals infrastructure.

I made a compat.h and compat.c to replace the missing function calls with dummy ones like you did with your TCL experiment.

I also 'borrowed' the GNU/Linux versions of strnlen.c and getopt.c for now. There's a CDL option to supply comandline args to main{}, so maybe an ecos implementation of getopt() might be useful to have in a compatability package or something.

No doubt that it will come off the rails as soon as it hits one of the dummy calls, but hopefully with some more work it will be able to load a file off a jffs2 partition, edit it, and save it back.

Steven


[-- Attachment #2: compat.c --]
[-- Type: text/plain, Size: 880 bytes --]

#include <errno.h>
#include "compat.h"

int lstat(const char* f,struct stat* b){return stat(f,b);}
mode_t umask(mode_t mask){errno=ENOTSUP;return(-1);}
int execvp(){errno=ENOTSUP;return(-1);}
int execlp(const char *file, const char *arg, ...){errno=ENOTSUP;return(-1);}
pid_t fork(){errno=ENOTSUP;return(-1);}
//kill(){errno=ENOTSUP;return(-1);}
pid_t wait(int status){CYG_UNUSED_PARAM(int,status);errno=ENOTSUP;return(-1);}
int pipe(int pipefd[2]){errno=ENOTSUP;return(-1);}
char *mktemp(char *template){errno=ENOTSUP;return(0);}
int mkstemp(char *template){errno=ENOTSUP;return(0);}
struct passwd *getpwnam(){errno=ENOTSUP;return(0);}
struct passwd *getpwent(void){errno=ENOTSUP;return(0);}
struct passwd *getpwuid(uid_t uid){CYG_UNUSED_PARAM(uid_t,uid);errno=ENOTSUP; return(0);}
pid_t waitpid(pid_t pid, int *status, int options){errno=ENOTSUP;return(-1);};
void endpwent(){}

[-- Attachment #3: compat.h --]
[-- Type: text/plain, Size: 1223 bytes --]

#ifndef CYGONCE_NANO_COMPAT_H
#define CYGONCE_NANO_COMPAT_H
//#include <errno.h>
//#include <unistd.h>
#include <sys/stat.h>
struct passwd
{
};

#define P_tmpdir "/tmp"
#define   SIGCONT         18      /* Continue (POSIX).  */
#define SIGSTOP         19      /* Stop, unblockable (POSIX).  */
#define   SIGTSTP         20      /* Keyboard stop (POSIX).  */

int getopt(int argc, char * const argv[], const char *optstring);
size_t strnlen(const char *s, size_t maxlen);
extern char *optarg;
extern int optind, opterr, optopt;

//isascii(c){return((unsigned)(c)<=0200);}
int lstat(const char* f,struct stat* b);
mode_t umask(mode_t mask);

//readlink(p,b,s){int 
//n,f;if(f=open(p,1)==-1)return(-1);n=read(f,b,s);close(f);return(n);}
//char *mktemp(char *template);
int mkstemp(char *template);
//struct passwd *getpwnam(const char *name);
struct passwd *getpwent(void);
void endpwent(void);
struct passwd *getpwuid(uid_t uid);
//int execvp(const char *file, char *const argv[]);
int execlp(const char *file, const char *arg, ...);
pid_t fork(void);
//kill(){errno=ENOTSUP;return(-1);}
//pid_t wait(int *status);
pid_t waitpid(pid_t pid, int *status, int options);
//int pipe(int pipefd[2]);
#endif //CYGONCE_NANO_COMPAT_H

[-- Attachment #4: Type: text/plain, Size: 148 bytes --]

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Nano using PDCurses for eCos

[Sergei Gavrikov]

[-- Attachment #1: Type: text/plain, Size: 3750 bytes --]

Steven Clugston wrote:
> Sergei Gavrikov wrote:
> > If you just want to edit config files, it's better to implement own
> > editor, call it `femto' then :-)
> > http://en.wikipedia.org/wiki/SI_prefix
> > It seems for me that nano is enough bloat stuff for eCos targets.
> 
> Nano does far more than I realised and I agree that most of it is
> bloat for what I need like calls to external programs for spell checking
> and wide character and NLS support etc, but it's a choice between writing
> an editor from scratch or hacking nano to work by disabling the bloat
> using its existing config mechanism. I think the latter should take less
> time/effort. It's a shame there's not a #define NANO_FEMTO mode to strip
> off even more than NANY_TINY does.

[snip]

> Sergei Gavrikov wrote:

[snip]

> > When I will get a far progress with Nano/PDCurses I wil let you know.

Hello Steven,

Let moderators excuse this post. At first, I thought to write you
personally, but, then I thought, Why not here? More that one guy did
share own eCos shell here:
http://sourceware.org/ml/ecos-discuss/2006-10/msg00141.html

And it seemed for me that our thread is a shell-related thread. You told
about your needs to edit config files and yesterday I thought, Why do
not write and send e-mails using SMTP with nano like pine does it using
pico :-) A hundred of kb is not issue for PC target, for example.  What
curses program does at all? The curses program the mostly sleeps.  IMHO,
curses is not bad thing for eCos. So, why do not benefit our experience
with other eCosers? IMHO, TCL shell or PDCurses shell is more powerful
thing than just "while(1) cyg_io_read(...);" It is a portable way to
get the shells.

I have got some results on "femto". Today I put nano-2.0.9 under local
mercurial repository and cleaned up all my tweks.

(1) My "femto" config.h version and my tweaks let me build the nano
    without eCos POSIX package. Instead I add CYGPKG_LINUX_COMPAT (just
    to get headers). AFAIK, some eCos FS packages depend on it (e.g,
    it's possible to add FS_FAT, FS_JFFS2 without bloat CYGPKG_POSIX).
    My config.h let me build nano without Unix stuff at all
    
    arm-eabi-nm src/nano | grep -E "T (wait|fork|exec)"
    (no output)

    For my config I have got only 5 stubs: getpwuid(), getpwent(),
    endpwent(), umask() and mkstemp().

(2) About POSIX signals. There were exactly two functions in nano.c:
    signal_init() and do_suspend() for my config. Both of the functions
    depend on TOGGLE_SUSPEND_KEY and NANO_SUSPEND_KEY. But what is "^Z"
    for the eCos World? We can quite ignore those handlers.

(3) My eCos config was

    Template: default
    Added:
     CYGPKG_IO_FILEIO
     CYGPKG_FS_RAM
     CYGPKG_LINUX_COMPAT
     CYGPKG_PDCURSES
    Removed:
     CYGPKG_LIBM

    and it did let me get "The GNU nano" for ARM target as

   text	   data	    bss	    dec	    hex	filename
 123384	   1408	  24256	 149048	  24638	src/nano


(4) I add getopt.[ch] under compat directory. And the nano the above
    contains it. I think you are right: it would not bad to prepare
    something like CYGPKG_SERVICES_GNU_GETOPT. It seems for me that
    today's eCos license let us do it. And, perhaps, you are right when
    said about CDL against getopt.

(5) My tweaks of nano/src seem itself the very minimal (only nano.h,
    and nano.c were tweaked), look at an attached nano.diff.  It was
    surprise for me. But, now I do not know what will happen if someone
    will turn on an external spell checkeri, for example, in my config.h 

I attach all stuff (including used getopt) to get "femto" under eCos. I
will put all on bitbucked then. If anyone is interested in any CLI under
eCos, he/she can use our experience.

Regards,

Sergei

[-- Attachment #2: nano-ecos-hack.tar.gz --]
[-- Type: application/octet-stream, Size: 14315 bytes --]

[-- Attachment #3: Type: text/plain, Size: 148 bytes --]

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Nano using PDCurses for eCos

[Sergei Gavrikov]

Sergei Gavrikov wrote:
> Steven Clugston wrote:

[ snip ]

> > As I mentioned before, I thought I'd have a go at getting nano to work.
> > 
> > I've tried version 2.0.7 from:
> > http://savannah.gnu.org/projects/nano

Hello

Yesterday evening I did grab the latest stable nano sources (2.0.9) from
here http://www.nano-editor.org, and tried to build it. After configure
and some tweaks I got The GNU nano editor for the eCos i386linux target
just in 110K using those PDCurses package and dummy driver.

> If you just want to edit config files, it's better to implement own
> editor, call it `femto' then :-) http://en.wikipedia.org/wiki/SI_prefix
> It seems for me that nano is enough bloat stuff for eCos targets.
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
I should not have told about until a try. Steven, I have to finalize and
test the build (at the least using eCos FS_RAM), I have to add the real
terminal driver to PDCurses. I do hope this will occur the next weekend
and then I will share a full story somewere on the web (some guys are
nervious if they cannot read here only a digest about eCos kernel gears
and cpu parts :-) There are a few hints the below

I configured the nano as: --enable-tiny --disable-nls --disable-utf8; I
have a bit tweaked config.h to point on that which things are absent in
eCos, made a peace with Linux/GNU (pwd, mkstemp, signals, getopt) and I
got the executable. There is alone rule in my eCos Makefile

    ACTUAL_CFLAGS := $(subst , -DHAVE_CONFIG_H=1,$(ACTUAL_CFLAGS))


Sergei

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Nano using PDCurses for eCos

[Sergei Gavrikov]

[-- Attachment #1: Type: text/plain, Size: 4430 bytes --]

Steven Clugston wrote:
> 
> >> Steven Clugston wrote:
> >> I've been wanting to do a port of a minimal console text editor
> >> something like nano or pico to eCos to do on board editing of
> >> configuration files to be storred in flash. I've not had
> >> time to look
> >> into it yet, but as nano (as far as I remember) uses ncurses, this
> >> might be just the thing to get the ball rolling.
> 
> 
> > I brief check the latest nano sources at the least they have
> >
> > ~/repo/nano$ grep -li pdcurses *
> > BUGS
> > ChangeLog.pre-2.1
> > NEWS
> >
> > ~/repo/nano/src$ grep -i pdcurses *
> > winio.c:#ifdef PDCURSES
> >
> > Sergei
> 
> Hi Sergei
> 
> I've downloaded your PDCurses for eCos and I'm now able to have coloured Xmas trees and worms in minicom running from my board - thanks.
> 
> Can this be put into CVS now? Or will you keep it as a separate project?

Hi Steven

I'm not going to maintain PDCurses package for eCos as a separate
project. I have FSF agreement and I posted the announce in the
ecos-patches list via eCos bugzilla engine
http://ecos.sourceware.org/ml/ecos-patches/2009-04/msg00009.html
So, I must be a patient. And I hope after some discussion the package in
that or another look can be committed in CVS. As Andrew said, he did not
saw the copyright issues with PDCurses itself. So, I hope, that will
happen.

> As I mentioned before, I thought I'd have a go at getting nano to work.
> 
> I've tried version 2.0.7 from:
> http://savannah.gnu.org/projects/nano
> 
> All went well until it came up against some the Linux/unix calls:
> 
> ../src/files.c: In function `safe_tempfile':
> ../src/files.c:1156: `P_tmpdir' undeclared (first use in this function)
> ../src/files.c:1156: (Each undeclared identifier is reported only once
> ../src/files.c:1156: for each function it appears in.)
> ../src/files.c:1165: warning: implicit declaration of function `umask'
> ../src/files.c:1168: warning: implicit declaration of function `mkstemp'
> ../src/files.c: In function `write_file':
> ../src/files.c:1376: warning: implicit declaration of function `lstat'
> 
> Ok, so I created a compat.h to delcare P_tmpdir as "/tmp" and map lstat to stat in a similar way as you did for TCL a few years ago (BTW what happened with you TCl eCos port? Is it available anywhere?).

That was yet another weekend mechanic. I just ran John Osterhout Tcl 6.7
on my small foot-print target. A community right decided that Jim is a
more suitable for embedded. But, I liked that Tcl 6.7 experience.

> Now I've got a problem with termios, as even thought <termios.h> is included and present in the ecos install dir I still get errors:
> 
> powerpc-eabi-gcc -DHAVE_CONFIG_H -I"${BUILD_PATH}/eCos/install/include" -I"${BUILD_PATH}/src/nano/src" -I"src" -O3 -Wall -c -fmessage-length=0 -MMD -MP -MF"src/nano.d" -MT"src/nano.d" -o"src/nano.o" "../src/nano.c"
> ../src/nano.c: In function `finish':
> ../src/nano.c:583: warning: implicit declaration of function `tcsetattr'
> ../src/nano.c:583: `TCSANOW' undeclared (first use in this function)
> <snip>

Hm,

$ find $ECOS_REPOSITORY -type f -name \*.h -print0|xargs -0 grep -l TCSANOW
/home/sg/repo/devo/ecos/packages/isoinfra/current/include/termios.h

check `-I' pathes and eCos `ifdefs'.

> Anyway, that's my progress so far.
> 
> Also there's still the issue of no input support in the dummy pdcurses driver yet which makes nano a bit limited in use...

http://bitbucket.org/tickling/pdcurses4ecos/src/
http://bitbucket.org/tickling/pdcurses4ecos/src/tip/synth-gdb.txt


My first implementation for my own tests supported and input on the test
serial port just to interact with the demos. I did not use termios, but
eCos raw serial non-blocked i/o. I rejected those 3 functions from that
job (see an attachment). It was done just to interact with PDCurses
demos, no more. I used synthetic serial driver from Savin Zlobec, and
"nullmodem" from Jürgen Rinas, and minicom on pseudo tty. It worked. The
demos ran and on one my ARM target too.

IMHO

If you just want to edit config files, it's better to implement own
editor, call it `femto' then :-) http://en.wikipedia.org/wiki/SI_prefix
It seems for me that nano is enough bloat stuff for eCos targets. But
this my opinion, only.


Regards,

Sergei

> Regards,
> Steven
> 
> --
> Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
> and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[-- Attachment #2: snip.c --]
[-- Type: text/x-csrc, Size: 1997 bytes --]

static cyg_io_handle_t ser = NULL;

bool
PDC_check_key(void)
{
    cyg_uint32      len;
    cyg_serial_buf_info_t dev_buf_conf;

    len = sizeof(dev_buf_conf);
    cyg_io_get_config(ser, CYG_IO_GET_CONFIG_SERIAL_BUFFER_INFO,
                      &dev_buf_conf, &len);

    return (dev_buf_conf.rx_count > 0);
}

void
PDC_flushinp(void)
{
    cyg_io_get_config(ser, CYG_IO_GET_CONFIG_SERIAL_INPUT_FLUSH, NULL,
                      NULL);
}

// Note: 115200,n,8,1 did let me to collect the multi-bytes sequences on a
//       kbhit in such a simple form. It's ugly, but it worked.
int
PDC_get_key(void)
{
    cyg_uint32      buf;
    cyg_uint32      len = 4;

    cyg_io_read(ser, &buf, &len);

    // A very few support for PDCurses demos only
    switch (buf) {
        // ^[[ ... PC cursor movement keys
    case 0x415b1b:
        return KEY_UP;
    case 0x425b1b:
        return KEY_DOWN;
    case 0x435b1b:
        return KEY_RIGHT;
    case 0x445b1b:
        return KEY_LEFT;
        // Tuidemo wants 'Alt-X'
    case 0x581b:
    case 0x781b:
        return ALT_X;
    }
    // Nothing more
    return buf & 0xff;
}

static int      ttyline_set = false;
int
PDC_scr_open(int argc, char **argv)
{
    if (!ttyline_set) {
        cyg_uint32      wait = 0;
        cyg_uint32      len = 4;

        if (ENOERR != cyg_io_lookup(PDC_SERIAL_DEVICE, &ser))
            return ERR;
        if (ENOERR != cyg_io_set_config(ser,
                                        CYG_IO_SET_CONFIG_READ_BLOCKING,
                                        &wait, &len))
            return ERR;
        if (ENOERR != cyg_io_set_config(ser,
                                        CYG_IO_SET_CONFIG_WRITE_BLOCKING,
                                        &wait, &len))
            return ERR;

        ttyline_set = true;
    }

    SP = calloc(1, sizeof(SCREEN));

    if (!SP)
        return ERR;

    SP->cols = PDC_get_columns();
    SP->lines = PDC_get_rows();
    SP->cursrow = SP->curscol = 0;

    // ...
}


[-- Attachment #3: Type: text/plain, Size: 148 bytes --]

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Nano using PDCurses for eCos

[Steven Clugston]

>> Steven Clugston wrote:
>> I've been wanting to do a port of a minimal console text editor
>> something like nano or pico to eCos to do on board editing of
>> configuration files to be storred in flash. I've not had
>> time to look
>> into it yet, but as nano (as far as I remember) uses ncurses, this
>> might be just the thing to get the ball rolling.


> I brief check the latest nano sources at the least they have
>
> ~/repo/nano$ grep -li pdcurses *
> BUGS
> ChangeLog.pre-2.1
> NEWS
>
> ~/repo/nano/src$ grep -i pdcurses *
> winio.c:#ifdef PDCURSES
>
> Sergei

Hi Sergei

I've downloaded your PDCurses for eCos and I'm now able to have coloured Xmas trees and worms in minicom running from my board - thanks.

Can this be put into CVS now? Or will you keep it as a separate project?

As I mentioned before, I thought I'd have a go at getting nano to work.

I've tried version 2.0.7 from:
http://savannah.gnu.org/projects/nano

All went well until it came up against some the Linux/unix calls:

../src/files.c: In function `safe_tempfile':
../src/files.c:1156: `P_tmpdir' undeclared (first use in this function)
../src/files.c:1156: (Each undeclared identifier is reported only once
../src/files.c:1156: for each function it appears in.)
../src/files.c:1165: warning: implicit declaration of function `umask'
../src/files.c:1168: warning: implicit declaration of function `mkstemp'
../src/files.c: In function `write_file':
../src/files.c:1376: warning: implicit declaration of function `lstat'

Ok, so I created a compat.h to delcare P_tmpdir as "/tmp" and map lstat to stat in a similar way as you did for TCL a few years ago (BTW what happened with you TCl eCos port? Is it available anywhere?).

Now I've got a problem with termios, as even thought <termios.h> is included and present in the ecos install dir I still get errors:

powerpc-eabi-gcc -DHAVE_CONFIG_H -I"${BUILD_PATH}/eCos/install/include" -I"${BUILD_PATH}/src/nano/src" -I"src" -O3 -Wall -c -fmessage-length=0 -MMD -MP -MF"src/nano.d" -MT"src/nano.d" -o"src/nano.o" "../src/nano.c"
../src/nano.c: In function `finish':
../src/nano.c:583: warning: implicit declaration of function `tcsetattr'
../src/nano.c:583: `TCSANOW' undeclared (first use in this function)
<snip>

Anyway, that's my progress so far.

Also there's still the issue of no input support in the dummy pdcurses driver yet which makes nano a bit limited in use...

Regards,
Steven

--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss