[leon_sparc] Re: CRLF vs (just) LF

[Joel Sherrill]

Jarl Friis wrote:
> 
> Ralf Corsepius <corsepiu at faw.uni-ulm.de> writes:
> 
> > Am Don, 2002-01-17 um 13.27 schrieb Jarl Friis:
> > > Ralf Corsepius <corsepiu at faw.uni-ulm.de> writes:
> > >
> > > > Am Don, 2002-01-17 um 12.48 schrieb Jarl Friis:
> > > > > I just installed and ran the rtems-hello.c program.
> > > > >
> > > > > One thing that I find annoying is that the "\n" character in the
> > > > > printf string translates to CR LF (two bytes). I expected that it
> > > > > would as on any other relevant OS generate only LF. I doubt this is an
> > > > > RTEMS issue but related to the sparc-rtems-gcc version found on
> > > > > www.gaisler.com.
> > > > >
> > > > > Can someone direct of where to start to look to resolve this problem?
> > > > * termios, c.f. [I|O]CRNL, [I|O]NLCR etc.
> > >
> > > eh?, I think I kind of wanted a path to a file in the srouce tree?
> > > What are you trying to say here. I assume that we agree to LF==NL.
> > OK, apparently, I was too brief.
> >
> > With termios-supported console-drivers, automatic LF->CR-LF conversion
> > is a setable attribute of termios (cf. man tcgetattr, man tcsetattr).
> >
> > Therefore, you might want to read man termios (rsp. man termio on some
> > hosts), read RTEMS's termios.h, and check for the flags ICRNL, OCRNL,
> > INLCR, ONLCR, which are involved into processing of CR/LFs.
> 
> Oh yes... Things are clearing up now ;-) rtems is a posix kernel
> right? I'll have a look at the use of termios.
> 
> Thanks, so my rtsem hello world Version 2 will probably be one that
> prints the termnal settings :-)
> 
> > > > * console-drivers: AFAIK, some (non-termios-based) console drivers apply
> > > > LF->CRLF conversion (Note: These are console-drivers, not serial
> > > > drivers, being applicable for general serial IO purposes.)
> > > >
> > > > * Check the settings of your terminal.
> > >
> > > My terminal does not do that. it is an issue of the compiled
> > > rtems-hello program. the (rest) of tsim does not generate CRLF (only
> > > in one line (using perf), which I consider a bug. Any other app
> > > behaves as expected as well
> > This normally is a customizable option of terminal programs, similar to
> > baud-rates, parity-checking etc.
> >
> > [I haven't seen any terminal emulator which doesn't have this options.]
> >
> 
> I guess we are now talking about the terminal I run tsim from (on the
> host coputer), right?
> Assuming I pipe the output to a file I can be sure no interpretation
> by any terminal is done to the output from UART1 on LEON (i.e. tsim)
> or the messages generated by tsim itself.
> 
> That makes me wonder where the implementation of printf when making
> boot prom programs.
> 
> I was even wrong on my state on the hello-world program as boot prom,
> for each "\n" printf emits "\n\r" (note the non-standard order of the two
> bytes), further printf does not emit the two last bytes of a
> string. Issuing printf("Hello") just gives you a "Hel" on the
> output. Normally you don't notice since "\n" will be exactly the two
> ("\n\r") last bytes that are not emitted.
> 
> Well, where ever that printf-implementation comes from I gotta have a
> look on that, actually I should not, making a boot prom I probably
> should implement my own.

The rtems printk() routine is pretty simple and shouldn't be difficult
to use in a bootloader.  Some BSPs build the bootloader side-by-side
the BSP, carefully splitting files to share things.

> > Ralf.
> 
> Thanks.
> 
> Jarl

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
   Support Available             (256) 722-9985