Re: [gforth] problem with gforth7 and serial access

[Bernd Paysan]

Am Donnerstag, 5. April 2012, 21:01:13 schrieb kibler:
> Our group will do more testing but as of now I can use it again.
> I might add that I was trying to use "pipe" functions and it would
> read as well, but would not give me a valid key?-file indication.

You mean with open-pipe?  I checked pipes (shell redirect), files, serial and 
console IO, and it all works.  open-pipe doesn't seem to work.  Well, I 
replaced select with poll, and now key?-file works on open-pipe fds, too.

I'm quite pissed about that, it essentially means that select() is broken 
beyond repair on Linux.

> I suspect it is the select is failing, as it was for tty, but at this
> point I can not be truly sure of that. When looking at the engine
> code I get very concerned about the "select.c" module, and think
> it might be what is called by io.c and not the actual system "select".

Forget this select.c module, it is not even checked in configure.  It's only 
there for MS-DOS, which we no longer test (through lack of demand ;-).  It 
will never ever, even by accident, get used.

You can call "strace gforth" to see which syscalls actually are performed.

> If that is true, it would account for the failure, since select.c only
> adjusts the timeout values and nothing more.

Yes, that's what the select replacement on DOS is supposed to do.

-- 
Bernd Paysan
"If you want it done right, you have to do it yourself"
http://bernd-paysan.de/

signature.asc
Description: This is a digitally signed message part.